Update on my development of a "plan language"

I received a request to post an update on the "plan language" on which I was working and that I mentioned a few times (e.g., https://www.chess.com/forum/view/chess-openings/what-does-a-repertoire-even-look-like-1), so here is my promised thread.

This is a *huge* topic, where each subtopic could easily be its own thread, so I'll post a few topics and insights in each post, and if anyone is interested, they can keep the thread alive by asking questions or posting comments. I was reluctant to post anything at all on this topic because it is so huge, although if it is that huge, maybe an intermediate thread would be a good idea, before the topic becomes so extremely large that I simply give up on trying to tackle the entire topic.

Some of the subtopics that are very large are: the nature and structure of any language at all, insights I've gotten on openings as a result of my study, hierarchies of plans, nameable patterns, tradeoffs. Also, my plan language is still changing rapidly, so the words I'm using may well change.

Here is the start of some subtopics to get a discussion started...

(1) Event sequences

One of my first intents was to overcome a common problem of opening books, which is omission of the general plan behind moves. (Essentially I'm starting to write my own opening book, out of necessity, focused on my repertoire, which anybody else can do, too, and which I recommend.) However, because plans are typically hierarchical this structure can complicate the representation, and I'm still working on this issue. The best I have so far is a start on a representation method for temporarally ordered events, using the symbol "<=", and I'm tenatively going to switch to the symbol ":" to represent a subplan, since I'm already using ":" to represent subvariations.

For example, the following notation...

prepare-for Be2 <= prepare-for O-O

...means the current move (Qd1) is done to prepare for the later move Be2 (by moving the queen out of the way), which in turn is done to prepare for the move O-O.

Below is a screen snap of my repertoire file where I use this notation. Note that the section name uses multiple names separated by ":" where the more basic opening names (especially official names of early opening moves) are at the beginning, and the more detailed variations are at the end.

(2) Variation names

Also note in the above excerpt that I'm using two different types of quotes: single quotes and double quotes. Quotes mean that those are my own names, not official names.

Most common are the double quotes (e.g., "Augstein Defense."), which I usually take from the names of players from online opening databases, and the single quotes (e.g., 'Calthrop Attack.') are also my own names but more historically backed up by the book "Oxford Encyclopedia of Chess Games, Volume 1, 1485-1866," usually by selecting the earliest recorded game with that move.

(3) Temporally spread events

One of my new developments has been to use bracket-like symbols (viz., "<...>") for events that are spread out over time. For example, "<pawn-storm", which shows only the left bracket to indicate that the event has started, does not typically have a closing bracket because a pawn storm typically continues indefinitely. That example is shown in the above screen snap, too.

There exist temporally spread events that are shorter and typically have a concluding move, though, such as the event I named "natgeadan" = knight advance to get exchanged and drive away knight, which typically (but not necessarily) involves the moves Ne5 ...Nxe5 dxe5 ...Nd5. Below is a screen snap of the start and end of that shorter "natgeadan" sequence. Note that this example contains an even shorter temporal sequence: "feex", which is my abbreviation for "fast, even exchange," which is extremely common in all openings, which is why I used an abbreviation (due to Zipf's Law).

(4) Tactical terminology

By far the most common usage of plan language is to explain what each single move does, which at least a few chess game books feature, unlike most opening books. These are also the easiest explanations to understand, since most commonly moves are done for tactical reasons. This doesn't quite prove Teichmann's claim that "chess is 95% tactics" but after studying enough books moves it does provide insight into the nature of "best" moves,  namely that they usually accomplish several good things at once.

(5) Personal repertoire comments

Note that all comments in my repertoire are done within braces ("{...}"), since this is legal PGN notation that allows me to cut-and-paste openings or games into chess.com diagrams with relative ease (at least after I move Black's move numbers like "11..."). There is a clever little trick I use here: by accident I discovered that chess.com's PGN player will not display the first of multiple sets of braces, therefore I put personal repertoire notes within a separate, first set of braces that I know will not be visible to the public. After all, a repertoire is essentially a battle plan that you don't want to disclose to the enemy. (Of course I made a small exception when I posted this diagram.) This allows me to cut-and-paste without fear of giving away the lines I personally choose to play, which are marked by "pt" (= pro tempore). I also try to put in a comment as to why I decided on that particular move, via a "due-to" comment. See the screen snap below.

Also note in the example above that it shows two possible transpositions that will arrive at that same position.

As I mentioned, this is all a huge topic, and I'll stop there for my first post in this thread.

[Please allow 30 minutes after this repertoire post appears for me to notice and fix any typos and inaccurate wordings.]

I was busy and couldn't follow up with more posts to this thread for several days. My development keeps rushing forward, though, and even some of the conventions I posted have already changed. Here are some more random subtopics...

(6) u-attack and s-attack

Since the term "attack" is very ambiguous, I distinguished between two different meanings of the word.

Per my convention a "u-attack" is an attack on a unit, and an "s-attack" is an attack on a square. By far the most pervasive of these two in my repertoire so far is "u-attack" partly since tactics are so dominant when describing the intents of moves.

For an example of a "u-attack," we're all familiar with the opening moves 1. e4 e5 2. Nf3, where White's knight attack's Black's e5-pawn. That's a "u-attack," and can be seen in my repertoire as such:

For an example where an "s-attack" is important, the following is a puzzle that involves an x-ray attack, where the critical theme is an attack on a square (in this case the h8-square) instead of an attack on a unit...

I'm still unhappy with use of the term "attack" at all. I'd prefer my own term "forcelight", and the distinction as to whether a unit is actually threatened versus merely covered is not sufficiently clear by my current terminology, but that's another huge area and an independent research project.

(7) types of u-attack and possible responses

As we all learned when we first learned how to play chess, there exist only three types of legal responses to escape a single check: (1) the king flees; (2) the king is shielded along the checking line, if possible; (3) the checking unit is captured.

In a double check there exists only the first option. Most of us probably haven't considered the types of responses for when a nonking unit is attacked however. I've listed the possible options below by considering all combinations. Some combinations are illegal, and some are even physically impossible.

----------

TYPES OF U-ATTACKS (in descending order of priority)

1. double check (double u-attack on king)

   dc-attack

   dcr = double check response

2. single check (u-attack on king)

   c-attack

   cr = check response

3. double attack on minion

   (?) dm-attack

   uar = unit attack response

4. single attack on minion

   (?) sm-attack

TYPES OF RESPONSES

(1) flee

   dcr-flee

   cr-flee

   uar-flee

   uar-flee-and-u-attack-attacker

   uar-flee-and-u-attack U

   uar-flee-and-capture U

(2) shield

   dcr-shield - impossible

   cr-shield

      of-type

         active

         passive

   uar-shield

(3) capture

   dcr-capture - impossible

   cr-capture

   uar-capture

(4) supersede (zwischenzug)

   dcr-supersede - illegal

   cr-supersede - illegal

   uar-supersede

(5) protect

   dcr-protect - illegal

   cr-protect - illegal

   uar-protect

   uar-protect-and-u-attack U

(6) ignore

   dcr-ignore - illegal

   cr-ignore - illegal

   uar-ignore

----------

Admittedly I'm not currently using some of this notation, such as the terms marked with "(?)", and I currently I prefer "zwischenzug" over "supersede", but this is likely to change eventually. At least it shows my train of thought.

(3) insight into book moves: fleeing is often accompanied by counterattacking

One insight I gained from studying the most popular moves from databases is that, as I already knew, the best moves do multiple good things at once, in particular when a unit must flee it is often repositioned such that it counterattacks as well. This is so common that I even created modified terms that note this behavior, as you can see in the above response list. This is obviously something you should do when playing non-book moves, too: try to find a response that counterattacks upon fleeing.

Below is one example from my repertoire that shows this situation via terms like "uar-flee-and-u-attack by-pawn". The most common situation is "u-attack" followed in the next ply by some form of response "uar-", which means "u-attack response." Common responses are "uar-flee" and "uar-protect".

(4) insight into book moves: time to complete development

Seventeen months ago some user asked an excellent question on this site: How long does it typically take to complete development in an opening? (https://www.chess.com/forum/view/general/how-many-moves-does-it-usually-take-you-to-complete-your-basic-development) I've been wondering about this ever since, and now with all this database study behind me I now know the approximate answer: 20 moves. This is double the answer that other people posted (10 moves). Apparently the opening books that tell you to develop as fast as possible aren't realistic: in real life numerous side attacks typically happen before both sides get castled and get all their units off their original squares.

A good example of this is seen in the above diagram of the Center Game: in the line shown it took White 25 moves to get completely developed, and it took Black 27 moves to get completely developed. Usually it is within 20 moves, give or take 5 moves, but this illustrates my point well.

This means that this type of notation for plans can also be used for academic and statistical purposes: by searching for keywords one can collect data on the nature of openings in general, or even on specific openings.

(5) plus and minus situations

I use keywords to indicate any situation that is of interest to me or others, not just whether a plan is involved.

For example, I have terminology for when pawns become doubled or isolated, since those characteristics are often critical for planning an attack. As a small refinement I began prefixing terms with "-" if they are generally considered negative situations. Like + and - signs on numbers however, I don't bother to put "+" since + is the default and it would be too annoying to have to write that extra symbol all the time. (Zipf's Law, again.)

Specifically, I put "-" prefixes on terms for doubled pawns and isolated pawns, for example, even though some people will debate whether an isolated pawn is always a liability. (It's *my* repertoire and language so I can codify my opinion if I want to! You can modify your own such repertoire and language according to your own opinion.)

For example, in my repertoire excerpt below you can see reference to the following terms with "-" in front of them:

-doubled-pawns

-isolated-pawn

(11) predominant outcome

A typical move database probably shows the statistical results of a single move by listing wins for White, wins for Black, and draws, as shown:

Since I am attempting to squeeze as much information as possible into a small space in my repertoire, I condense the results of a move into a single value I call the "predominant outcome." The predominant outcome is the highest valued outcome of the three mentioned values. For example, in the first line of outcomes of the move choices shown, [36.6% 43.8% 19.7%], I would represent this as "n 44%", which means that usually nobody ("n") wins,  44% of the time.

One drawback of this method is that the predominant outcome can flip-flop between White and Black winning as their values become close in value. Another drawback is that there can be ties. In the case of ties I simply combine the outcome letters. For example, if both White and Black win 50% of the games I would represent the predominant outcome as "wb 50%".

One drawback of this method is that it doesn't show the number of games used in the average, which means the predominant outcome value is not very useful, although on ends of lines in book my "eolib" notes serve as an inherent caution not to trust the percentage, which will always be 100% in that case.

(12) ends of lines

Move databases can end lines in various ways. The most common way is to end in a leaf node of the move tree, where the leaf is a complete game. My abbreviation for this case is "eolib" = end of line in book.

For example, in the above screen snap of a move database, the three least popular moves, shown at the bottom of the list, consist of specific games. For example, I would represent the bottom line here as...

5...Qd6 {eolib. w 100%.}

...or if I want to name that variation...

5...Qd6 {"Darouny Defense." eolib. w 100%.} 

The predominant outcome percentage of course must always be 100% on leaf nodes.

By the way, I round my percentages to the nearest integer, unlike some databases, though I determine the predominant outcome from the unrounded percentages.

Another way a line can end is when a person or computer plays a move that is not in book. I represent that case by "oob" = out of book.

For example, if the move played or considered were 6...h5, I would represent that as...

5...h5 {oob.}

A similar abbreviation I use is "oor" = out of repertoire, which is a line I have in my repertoire that I don't actually play, for some reason.

Another abbreviation I use is "omib", which is "only move in book.", as shown below in database form and in my notation convention...

Sometimes this happens because it is the only legal move, as in Black's recapture move #8 above, or it can be the only *logical* move, or it could just be that the database is nearing its leaf nodes there, meaning there is little data available,

which can also be seen in the above repertoire screen snap starting around move #15.

(13) insight into book moves: length of lines

Another insight into book moves is that the book lines rarely exceed very far beyond 20 moves, maybe up to 30 as the upper limit in my experience. As I mentioned before, the average time for either or both players to fully develop is about 20 moves, give or take 5 moves, so at that point the opening could be said to be over and the middlegame beginning, so there is less need for an opening database at that point. I suspect the very nature of the middlegame contributes to this sudden but natural paucity of data because the middlegame is the phase of the game with the highest complexity, implying the largest number of possibilities.

This can be seen in the above screen snap of my repertoire for the Berlin Ruy Lopez, which ends at the 27th move with two leaf nodes (single games).

Again, note that it took about 20 moves for each side to develop: Black at move #19, White at move #22.

(14) modifiers

In English grammar (in fact, in all natural languages), there exist words that function as modifiers (https://en.wikipedia.org/wiki/Grammatical_modifier). "Modifers" is a general term that describes a function, a function that can be done by adjectives, adverbs, prepositional phrases, or other.

We've seen these in diagrammed sentence in English class, where modifiers are the branches hanging off the bottom of the important words of a sentence....

I regard modifiers as describing a hierarchy. For example, one way I've incorporated modifiers in a fairly rigorous way in my planning language is by prepositional phrases.

For example, the above diagram modifies "games" with the adjective "video", so the set of games could be considered a superset and video games a subset. In my notation I would represent that as "games of-type video", although that is not a chess topic.

As an example of a chess topic, consider the concept of a pawn duo. In the classical school of chess in an e-pawn opening we try to establish a center pawn duo consisting of an e4-pawn and a d4-pawn.

In a d-pawn opening, however, since it's usually too difficult to play a safe e4, we have to settle for an off-center pawn duo consisting of a d4-pawn and a c4-pawn.

While these two types of duos are subsets of the general concept of a pawn duo, they can be distinguished in my notation by the modifier/preposition "of-type" as follows: 

duo of-type center

duo of-type off-center

I notate other types of duos, as well, which I might discuss later, but these will suffice for this example.

In the screen snap of my repertoire below, this terminology can be seen at move 2. c4...

One nice aspect of this notation is that if you forget my exact subtype syntactical name, or are unsure, you can just put "duo" and figure it out later, and refind the location by doing a Notepad search on the word "duo".

This is similar to how some people use the very ambiguous word "thing": "Put that metal thing on the extension thing into a horizontal postion."

(15) fixed terminology

As in virtually any language, including computer languages, there exists only a finite set of offical words that can be used. Such word lists in natural languages are commonly collected into books called dictionaries.

Therefore at the beginning of my repertoire I also have a dictionary of the terminology I use, marked as in standard dictionaries with "Noun" or "Preposition", etc., to make clear the part of speech to which that term corresponds.

That is also where I put definitions of those terms, as found in natural language dictionaries.

Originally I had collected words and their definitions into subject sections at the beginning of my repertoire file, but I was eventually forced into the dictionary convention due to overlap of concepts. For example, my term "nabafap" (= knight attacks bishop and frees a pawn) applies to a temporal sequence, a piece attacking another piece, and a pawn advance, therefore I used to keep a copy of the term in three different locations in the file. This had the expected drawback that if I modified my definition in one location I would often forget to update the definition in the other locations, as well. A dictionary that had all definitions in a single place fixed this problem nicely. 

[Please allow 30 minutes after this repertoire post appears for me to notice and fix any typos and inaccurate wordings.]

What else?

I'm going to be low on time for three days so I'll just post two topics tonight.

(16) two types of popularity measures

Notice that I'm using terms like "#1 pop" and "#2 pop" to mean 1st most popular move, 2nd most popular move, etc. In the "old days," meaning about a year ago when I started on my new type of repertoire text file, I used a more elaborate system that was more accurate. For example, at one time I listed the first main responses to 1. e4 as something like:

1. e4

   1...c5 {Sicilian Defense. 673034/1622553 = 41% pop.}

   1...e5 {King's Pawn Game. 382883/1622553 = 24% pop.}

   1...e6 {French Defense. 214969/1622553 = 13% pop.}

   1...c6 {Caro-Kann Defense. 114422/1622553 = 7% pop.}

   1...d6 {"Pribyl Defense." 71701/1622553 = 4% pop.}

   1...d5 {Scandinavian Defense. 59846/1622553 = 4% pop.}

   1...g6 {Robatsch Defense. 49189/1622553 = 3% pop.}

   1...Nf6 {Alekhine's Defense. 39145/1622553 = 2% pop.}

...

To get that total 1622553 I had to manually add up all the numbers of games used in the database for each of those 1st move responses, by hand. Those numbers are already in the database, all on the same page, so it wasn't hard, just time-consuming, especially for early opening moves because there are typically many possible responses early in the game. Then I just divided each response's number of games by that same total to get the relative percentage of the time that that move was used at that point in the game.

Nowadays I don't have the time to do those calculations, but that's the more accurate and preferable way to get a number that gives a quick feel for move popularity. Occasionally I recopy those percentages into my new repertoire to use both the popularity ranking and popularity percentage such as....

1. e4

   1...c5 {Sicilian Defense. #1 pop = 673034/1622553 = 41%.}

   1...e5 {King's Pawn Game. #2 pop = 382883/1622553 = 24%.}

   1...e6 {French Defense. #3 pop = 214969/1622553 = 13%.}

   1...c6 {Caro-Kann Defense. #4 pop = 114422/1622553 = 7%.}

   1...d6 {"Pribyl Defense." #5 pop = 71701/1622553 = 4%.}

   1...d5 {Scandinavian Defense. #6 pop = 59846/1622553 = 4%.}

   1...g6 {Robatsch Defense. #7 pop = 49189/1622553 = 3%.}

   1...Nf6 {Alekhine's Defense. #8 pop = 39145/1622553 = 2%.}

...

There are a number of questions and implications that arise when I examine how rapidly those popularity percentages decline as rank increases, but I won't get into that discussion here. The main way I use such percentages now is just as a quick estimate of which move choices to include in my repertoire: I stop including moves if they have popularity of 50% or less when compared to the most popular move.

(17) the number of participants in a sentence

In English grammar class we're taught about subject, direct object, and indirect object. That means the concept that is to be conveyed consists of three things connected by an action (implemented as the verb). I haven't yet found a term that describes the type of thing to which those three things generalize, so I'll call them "participants."

Many people are unaware of a striking grammatical fact: virtually every well-formed Engish sentence falls into one of exactly ten sentence patterns (https://webapps.towson.edu/ows/sentpatt.htm). In fact, I'm fairly sure all European languages have this same attribute of exactly ten sentence patterns.

Only sentence patterns #8 and #10 (see link in previous sentence) have all three participants. One natural question to ask is "What happens if more than three participants are involved?"

If we're designing a language that is to handle complicated situations well (yes, I'm talking about chess here!), presumably we need a very general grammar that will handle any number of participants, not just three.

One insight into handle such multiple participant situations is to ask how English does it. The answer: with prepositions.

For example, using sentence pattern #8 we can unambiguously describe the transfer of an object from one person to another, which involves three participants:

"John gave Mary the house."

participant #1: John

participant #2: Mary

participant #3: house

As an example of a concept involving four participants, consider that most real estate transactions involve a real estate agent because the average person does not know enough about the legalities of real property transfer. To include a real estate agent in this sentence we would say:

"John gave Mary the house through agent Emily."

The preposition "through" with the function "agent" together unambiguously describe how the fourth participant relates to the concept.

participant #1: John

participant #2: Mary

participant #3: house

participant #4: Emily

Now let's relate this to chess. An example of a 3-participant situation in chess is a unit protecting another unit that is under attack by an opponent's unit:

participant #1: unit under attack

participant #2: unit protecting the attacked unit

participant #3: unit attacking the unit

If that protecting unit is overworked, however, it is protecting yet another unit (or at least square), and that other unit becomes the 4th participant:

participant #1: unit under attack

participant #2: unit protecting the attacked unit

participant #3: unit attacking the unit

participant #4: another unit under attack protected by participant #2

Therefore we can see that chess gets at least as complicated as common situations in life. I'll post even more complicated examples next time, involving even more participants, if nobody can think of such an example before then.

I decided to handle the situation in roughly the same way that English handles them: use of prepositions.

For example, as White we're very used to the concept of "kicking" or "poking" or "biffing" a bishop via P-R3 (a3 or h3) to evict a bishop that is pinning our knight at B3 (c3 or f3).

Currently I'm using the term "biff" to apply only to kicking *bishops*, not knights or other outposted pieces at N4 (b4 or g4).

There exist a couple details of that situation that interest me that I want to document as a pattern: whether the biff comes immediately (versus delayed by more than one move), and whether it is followed by P-N4 to further evict that bishop, which I call a "complete" biff. Using prepositions marked with their meaning (remember the word "agent" in the above example, which clarified the exact meaning of the ambiguous preposition "through"), I can represent a biffing situation as something like...

6...Bg4

7. a3 {biff of-completeness incomplete of-timing immediate.}

In this way two modifiers can be attached to a situation, and it doesn't even matter much the order in which the prepositional phrases appear, since the meaning is not dependent on the order of prepositional phrases. This particular example uses only one participant, but the same method can be used to handle situations with any number of participants. In this way my plan language should be capable of easily representing situations of arbitrary complexity.

That's about all I have time to post tonight. I'll be back in three days.

[Please allow 30 minutes after this repertoire post appears for me to notice and fix any typos and inaccurate wordings.]

It turns out I have a couple hours before work, so here are a couple more topics...

(18) types of duos

I mentioned different types of pawn duos before, so for those interested, here are the types of pawn duos of which I am aware, and whose types I have named in my repertoire.

Since acquiring a pawn duo in the early opening is the main goal of the classical school of chess, this is an important topic, although I haven't yet figured out if and how I should represent this as a plan rather than as an after-the-fact pattern that occurred in a game. I haven't seen any book or other source mention another fact, though I mention this nowadays if I'm teaching someone the main ideas behind chess openings: the pawn duo (especially with the two knights at B3) is the ideal opening setup, and although your opponent usually thwarts your plan to achieve this ideal setup by trading off one of those duo pawns, you can still achieve almost the same position with one of the pawns missing. That's why I still call this type a "remainder" duo despite it having only one pawn: its intent is to be a pawn duo but with an invisible member of the duo.

By the way, I changed these names slightly after I started this thread, and I'm still playing with the names, so old screen snaps posted here might differ slightly in notation.

Here are some of the most common types of duos I've encountered, though I'm sure there exist more types:

(a)

(b)

(c)

(d)

(e)

(19) 5-participant concepts

I'll add to my previous post's discussion of concepts that involve several participants as a lead-in to a later discussion of various ways to handle numerous parameters in a complex concept. My example below shows that 5-participant concepts are reasonably common in real life and in chess.

Real estate example:

Because real estate involves exchanges of such expensive property, a situation that would ordinarily involve high risk of fraud, the standard practice is to use an escrow company as part of the process, in addition to the real estate agent. Therefore the escrow company becomes a fifth, integral part of the concept of a real estate transaction:

"John sold Mary the house through agent Emily through escrow Ngoc Thuy's Escrow."

participant #1: seller John

participant #2: buyer Mary

participant #3: property house

participant #4: agent Emily

participant #5: escrow Ngoc Thuy's Escrow

Chess example: 

"An overworked Black knight is guarding both a bishop and a pawn while White's knight attacks that pawn, and White's roook attacks that bishop."

participant #1: defending Black knight

participant #2: defended Black bishop

participant #3: defended Black pawn

participant #4: attacking White knight

participant #5: attacking White rook

It should be clear that such complexity can be arbitrarily extended. For example, in the real estate example the escrow company employees might speak only Vietnamese and therefore an interpreter might become an integral part of a 6-participant system. Likewise, in the chess example the rook capturing the bishop might also do so with check, resulting in a skewer of a queen behind the king. The queen would then be a 6th participant in the situation.

[Please allow 30 minutes after this repertoire post appears for me to notice and fix any typos and inaccurate wordings.]

(20) u-forcelighting versus u-threatening

As of yesterday I did a global change in my file of one frequent word in my terminology: I changed "u-attack" to "u-threaten", which in turn meant that my prefix "uar-" (u-attack response) changed to "utr-" (u-threat response). That's one problem with reporting on a work in progress. I may or may not retroactively update my earlier posts that mention that term. At least it's only one word, and understandable in either case. Note that I complained about that ambiguous word "attack" before, and yesterday I finally came up with a method I like that starts to disambiguate the term, which I'll describe here.

I believe one of the reasons humans learn quickly and intuitively how chess pieces move is that we already know from real life examples of things that function like that. This is particularly true of light, which travels in a straight line and stops at an obstruction.

I claim this is exactly how we understand so easily how sliding chess pieces work: if we were to put a flashlight, spotlight, or laser pointer at the base of a sliding piece and pointed in the legal direction of that piece's travel, the light hitting the squares would indicate exactly to which squares that piece could and couldn't travel. The only ambiguity would be at the final obstruction consisting of another unit: if it's our unit then we can't land there, if it's the opponent's unit then we can. This was my motivation for the term "forcelight": lines of force like magnetic force lines emanate from each unit and act like spot lights that indicate where that unit can travel. Force line + spotlight = forcelight. X-ray attacks are a clear-cut extension where instead of light rays we use x-rays to model the influence of those force lines emanating from chess units.

There is clearly a difference between forcelighting a unit and threatening that unit. For example, a White bishop aimed at a hanging Black knight threatens win of material by capturing that knight, but a mere pawn push that protects that knight typically negates that win-of-material threat immediately: the knight is still forcelit, but the threat of a win of material no longer exists.

Forcelighting is usually not significant enough to document for every move, especially since a moved unit usually forcelights several squares and units at the same time, which is too much information for a human to keep track of. However, piece placement is often chosen to forcelight pawns, pieces, or squares that are known to be common weak spots, especially for certain openings. For example, the h7-pawn in the French Defense is often the target of the Classical Bishop Sacrifice, and the f7-square is often the target of a similar sacrifice in Philidor's Defense, which explains why Bd3 is a standard developing move for White in the French Defense, and why Bc4 is a standard developing move for White in Philidor's Defense: history and experience tells us that those squares in those openings tend to be particularly promising. My recommendation and personal practice is therefore to document particularly important forcelit units or squares. Below are two games that show such important forcelighting, and the tactical wins that resulted from poor defense by Black in those positions.

(21) pseudo u-threats

Until I distinguished between "u-threaten" and "u-forcelight" I considered White's standard attack on Black's e5-pawn via 1. e4 e5 2. Nf3 to be a u-attack or a u-threat: a threat of win-of-material via Nxe5. However, more careful consideration shows that the situation is not that simple and that the threat is an empty one with respect to material: Black can recover that e5-pawn even if White steals it after Black plays carelessly for one move. For example:

For that reason I no longer describe that particular opening threat as a "u-threat" but rather as a "pseudo u-threat", as I documented in the three games above.

"Pseudo u-threats" can arise for other reasons, as well. For example, a common beginner's mistake is to biff a pinning bishop without noticing that if he really captured that bishop then he would lose a rook. Here's a made-up example of that, similar to one game I saw at our chess club recently:

Another example might be where a player about to be mated on the next move threatens a piece instead of preventing the mate.

As I mentioned before, accurate evaluation as to whether a forcelit unit will actually result in a win of material is extremely difficult to determine in every case. Gambits are an extreme example where we still don't know in some cases whether eventual, safe recovery of the gambited material is possible. Fortunately, not all subcases of win-of-material threats need to be specified in my plan language: my preposition "of-type" to specify the outcome does not need to be used at all. Simply writing "u-threat" will suffice even in cases with unknown theoretical outcome.

[Please allow 30 minutes after this repertoire post appears for me to notice and fix any typos and inaccurate wordings.]

(22) damped versus undamped strategies

I've heard of the philosophy of "win at all costs", which usually entails playing a highly tactical game, and/or possibly including taking extra risks, in order to ensure a win of the game instead of settling for a draw. However, most of us have also heard of the phrase "pressing too hard for a win," which results in losing a game that could have been easily drawn. Here is how I view these viewpoints in combination...

As in financial investments, I believe chess follows the general behavior of "high risk, high reward" systems: that an inherent tradeoff is built into the systems such that one value mathematically increases while the other value decreases. Since that type of phenomenon can be plotted in only two dimensions, that implies that by controlling only one variable (the "independent variable"), one can produce any possible allowable combination of those two values.

As I see it, highly tactical openings such as the Sicilian Defense are near the high risk end of the spectrum, and produce many decisive games, whereas low tactical openings such as the Exchange French are near the low risk end of the spectrum, and produce many draws.

An engineering analogy is that of damped versus undamped systems: an undamped system, which is a wildly, randomly fluctuating system will present frequent and unusually high peaks on both sides of the horizontal neutral line.

If a high peak represents a winning position for White, and a low peak represents a winning position for Black, then altogether there exist frequent winning chances for each side. (High risk, high reward.) The catch is that to exploit that peak, you must have the ability to spot a peak that wins for you, and to be able to keep up the pressure without lapse until your win is clear-cut.

At the other end of the spectrum, a damped system will lack extreme peaks altogether, which in a chess game represents an opening where neither side risks losing, but also neither side is likely to gain a winning advantage. (Low risk, low reward.)

Altogether this means conceptually that a single parameter describes the degree of tactics for a given opening. For the time being, let's assume that there exists only one extreme or the other: damped versus undamped.

With this assumption we can create two keywords that describe the type of game we desire to have:

damped-strategy

undamped-strategy

Possibly the word "strategy" should be replaced by the word "philosophy" in these terms: I'm not too choosy about the noun used at this point, only its meaning. With this fundamental distinction of two disparate strategies in place, many parts of the planning process automatically fall into place as described next. Remember that earlier in this thread I considered using a colon (":") to represent plan hierarchies. Well, this week I've decided to do that, even though I'm using such a colon to mean specific-to-general in the case of plans, whereas in the names of openings I use it in the other direction to mean general-to-specific. The context of the colon will be obvious, so no confusion should result.

Now consider how the choice of one of these two strategies affects move selection...

Suppose you're playing Black against 1. e4 and you want to follow the undamped-strategy. The top ten most popular responses are:

1. e4

   1...c5 {Sicilian Defense, q.v. #1 pop. w 36%. adod-opening:asymmetry-strategy:undamped-strategy.}

   1...e5 {Double King's Pawn Game (as White), q.v. #2 pop. symmetry-strategy:damped-strategy. w 41%.}

   1...e6 {French Defense, q.v. #3 pop. w 39%.}

   1...c6 {Caro-Kann Defense, q.v. #4 pop. w 37%.}

   1...d6 {"Pribyl Defense," q.v. #5 pop. w 41%.}

   1...d5 {Scandinavian Defense, q.v. #6 pop. w 41%.}

   1...g6 {Modern Defense, q.v. #7 pop. w 37%.}

   1...Nf6 {Alekhine's Defense, q.v. #8 pop. w 38%.}

   1...Nc6 {Nimzovich Defense, q.v. #9 pop. w 41%.}

   1...b6 {Owen Defence, q.v. #10 pop. w 47%.}

Using my new convention, to determine a move to match your undamped-strategy, you merely need to note which moves mention that undamped-strategy, in this case the Sicilian Defense (1...c5).

I've constructed a very general hierarchy of strategies that in this section looks like:

undamped-strategy

   asymmetry-strategy

      adod-opening

      half-open-opening

This means the undamped-strategy has two levels of substrategies. That abbreviation "adod" I'm using means "asymmetrical discouragement of duo." Yes, I have a lot of funky-looking names like that, many of which I'll describe later for fun and for knowledge.

If White instead played 1. d4, my applicable repertoire (cleaned up a little) looks like:

1. d4

   1...Nf6 {"Gligoric Defense," q.v. #1 pop. w 37%. develop. prevent e4.}

   1...d5 {#2 pop. w 40%. symmetry-strategy:damped-strategy.}

   1...e6 {#3 pop. w 39%.}

   1...f5 {Dutch Defense. #4 pop. w 40%. adod-opening:asymmetry-strategy:undamped-strategy.}

   1...g6 {Modern Defense. #5 pop. w 38%.}

   1...d6 {#6 pop. w 37%.}

   1...c5 {Old Benoni Defense. #7 pop. w 40%.}

   1...c6 {#8 pop. w 47%.}

   1...Nc6 {Lundin Defense. #9 pop. w 44%.}

   1...b6 {English Defense. #10 pop. w 45%.}

Again, there is marked a move indicative of conforming to the undamped-strategy: the Dutch Defense (1...f5), and for the same reason: "adod". (This is a brand new list that I haven't yet fully annotated, so forgive my ignoring other possibilities for the time being.)

With this system of strategy hierarchies my plan language can now connect moves to general strategies, although of course there is much more to say and to develop.

(23) an example of a temporal sequence: nabafap

Finding patterns in chess is fun, especially temporal patterns, especially when trying to give them names that are short, pronouncable, and memorable. One funky example is my term "nabafap", which means "knight attacks bishop and frees a pawn." By doing a search for this term in my repertoire file I can quickly find numerous games that used that pattern, including from openings in my file that I don't actually play, but maybe included just for general study. Below are three examples of this, which we've probably all seen in the Ruy Lopez. Since it is a temporal pattern, I mark its beginning and end with angled brackets, which I don't need to put on non-temporal pattens.

This idea of this move is to do a couple good things at once: attack a bishop while freeing the c-pawn so that it can advance to help control the center and gain space, so this sequence of moves is found in multiple openings.

[9-21-16: I removed the diagram here since I had overlooked a transposition.]

[Please allow 30 minutes after this repertoire post appears for me to notice and fix any typos and inaccurate wordings.]

(24) adod openings

My last post mentioned that I describe certain openings as "adod openings", so I thought I'd explain this a little more, and give statistical proof that such openings tend to produce decisive games rather than draws.

adod = asymmetrical discouragement of duo. This means that White's initial 2-square pawn advance is greeted by a Black 2-square pawn advance as well, positioned such that if White carries out his plan of completing his pawn duo (center or off-center), White will be forced to lose a tempo after the resulting immediate pawn trade, and sometimes even an entire pawn that might not be recoverable. There exist only four such openings if we limit our opening moves to the monarch pawns (d4-pawn and e-pawn) and bishop pawns (c-pawn and f-pawn). Here are those openings, shown here with the threatened loss of tempo ("lot") or loss of material ("lom") for White:

Now let's look at the statistics at Black's first response, for each adod-opening as compared to its counterpart symmetrical-strategy:

1. e4 c5 {Silician Defense. w 36%. n 30%. adod-opening.}

1. e4 e5 {Double King's Pawn Game. w 41%. n 30%. symmetry-strategy.}

Draw rate stays the same with symmetry-strategy.

1. d4 f5 {Dutch Defense. w 40%. n 29%. adod-opening.}

1. d4 d5 {Queen's Pawn Game. w 40%. n 34%. symmetry-strategy.}

Draw rate increases 5% with symmetry-strategy.

1. c4 {English Opening.} e5 {"Timman Defense." w 39%. n 31%. adod-opening.}

1. c4 {English Opening.} c5 {Symmetrical Variation. n 37%. symmetry-strategy.}

Draw rate increases 6% with symmetry-strategy.

1. f4 {Bird's Opening.} d5 {"Cramling Defense." b 40%. n 26%. adod-opening.}

1. f4 {Bird's Opening.} f5 {'Fraser Defense.' w 44%. n 26%. symmetry-strategy.}

Draw rate stays the same with symmetry-strategy.

The statistics aren't overwhelmingly convincing, but note that in no case did the outcome become *less* drawish when symmetry-strategy was used, and half the time the draw rate did increase by around 5-6% when symmetry-strategy was used. That's a definite indication that my claim is true: that adod openings tend to be decisive, whereas symmetrical openings tend to be drawish.

(25) an example of a temporal sequence: stagobee

Here's another temporal sequence whose existence will make you aware of a common and important opening pattern that no book explicitly teaches as a temporal pattern, as far as I know, if you're not already aware of it. I originally called it "staggered-double-bishop-development-due-to-monarch-door", but since this is really talking about staggered develop of bishops (Bs), I jokingly called it "stagobee" for short (Zipf's Law again), referring to the old song "Stagolee." The term "monarch door" is another term of mine that refers to one of the pawns in front of the monarchs (king or queen).

The idea is that in many games Black has lost one of his central pawns, and his remaining central pawn cannot advance more than one square. Since that pawn must block one of Black's bishops from developing beyond the second rank, no matter where the pawn is, Black's goal is to develop each bishop as it becomes free. First Black develops the bishop that is free while the pawn is unadvanced, then he advances that pawn one square (the monarch door opens), whereupon he develops the other bishop. This procedure must span at least three moves, so it's a temporal sequence. Below are three examples. Most commonly this is seen in the Scandinavian Defense, but it is found in other openings, as well.

 [Please allow 30 minutes after this repertoire post appears for me to notice and fix any typos and inaccurate wordings.]

(26) substrategies of damped and undamped strategies

Earlier I listed a few substrategies for the undamped-strategy. Below is a more complete list of my hierarchical list of substrategies for both damped and undamped strategies. 

This is how my list stands today, though it is still undergoing changes.

damped-strategy

   symmetry-strategy

   low-total-material

      queens-off-board

   weakly-tactical

      castled-same-side

   high-forcingness

      fast-positional-development

      few-open-lines

undamped-strategy

   asymmetry-strategy

      adod-opening

      half-open-opening

   high-total-material

      queens-on-board

   strongly-tactical

      castled-opposite-sides

   low-forcingness

      slow-positional-development

      many-open-lines

Some of my concerns and therefore some potential additions or modifications for this list are:

(1) "Plan obviousness" might be another factor, though for the time being I'm assuming that symmetrical considerations are the primary influence on how obvious plans are.

(2) Being castled on opposite sides might not contribute much to tactics when no opposite side pawn storms occur. I haven't yet figured out how best to merge both these concepts. Similarly, simultaneous pawn storms (especially on the kingside where both kings are castled) can get very tactical because both sides are voluntarily exposing their kings.

(3) There might be a more formal term for my concept "half-open-opening", by which I mean Black defenses that begin with only a single square pawn advance: the French, Caro-Kann, and Pribyl/Pirc defenses. I'll post this question elsewhere.

Since symmetry is so obvious to detect, let's look deeper at the contribution of symmetry on drawishness, via statistics. Since my statistics above when looking at a single symmetrical pawn response weren't strongly convincing, let's look at *two* consecutive symmetrical pawn responses, and compare them to the main lines of those opening moves...

Fully symmetrical line:

1. e4 e5 {w 41%. n 30%.}

1. e4 e5 2. Nf3 Nf6 {w 40%. n 36%.}

difference in draw rate = 36% - 30% = +6%

Main line:

1. e4 c5 2. Nf3 d6 {w 37%. n 29%.}

1. e4 e5 2. Nf3 Nf6 {w 40%. n 36%.}

difference in draw rate = 36% - 29% = +7%

Conclusion:

2 moves of symmetry increased the draw rate 6% compared to that of 1 move of symmetry.

2 moves of symmetry increased the draw rate 7% compared to that of main line.

Fully symmetrical line:

1. d4 d5 {w 40%. n 34%.}

1. d4 d5 2. c4 c5 {w 52%. n 28%.}

difference in draw rate = 28% - 34% = -6%

Main line:

1. d4 Nf6 2. c4 e6 {w 35%. n 39%.}

1. d4 d5 2. c4 c5 {w 52%. n 28%.}

difference in draw rate = 28% - 39% = -11%

Conclusion:

2 moves of symmetry decreased the draw rate 6% compared to that of 1 move of symmetry.

2 moves of symmetry decreased the draw rate 11% compared to that of main line.

Fully symmetrical line:

1. c4 c5 {w 37%. n 37%.}

1. c4 c5 2. Nc3 Nc6 {w 37%. n 36%.}

difference in draw rate =  36% - 37% = -1%

Main line:

1. c4 Nf6 2. Nc3 g6 {w 42%. n 29%.}

1. c4 c5 2. Nc3 Nc6 {w 37%. n 36%.}

difference in draw rate = 36% - 29% = +7%

Conclusion:

2 moves of symmetry decreased the draw rate 1% compared to that of 1 move of symmetry.

2 moves of symmetry increased the draw rate 7% compared to that of main line.

Fully symmetrical line:

1. f4 f5 {w 44%. n 26%.}

1. f4 f5 2. Nf3 Nf6 {w 38%. n 33%.}

difference in draw rate = 33% - 26% = +7%

Main line:

1. f4 d5 2. Nf3 Nf6 {w 35%. n 27%.}

1. f4 f5 2. Nf3 Nf6 {w 38%. n 33%.}

difference in draw rate = 33% - 27% = +6%

Conclusion:

2 moves of symmetry increased the draw rate 7% compared to that of 1 move of symmetry.

2 moves of symmetry increased the draw rate 6% compared to that of main line.

Grand conclusion, on the average...

Across all the opening moves shown, adding a 2nd symmetrical move to the 1st symmetrical move increases the draw rate 6/4 = 2%.

Across all the opening moves shown, adding a 2nd symmetrical move compared to the main line increases the draw rate 9/4 = 2%.

These are not very convincing statistics. However, note that the fully symmetrical d4 line seems anomalous since it has all negative values, probably since White's winning percentage increased so much, probably since that defense isn't very solid. If we leave out that anomalous line and average the others, we get...

Across all the non-d4 opening lines shown, adding a 2nd symmetrical move to the 1st symmetrical move increases the draw rate 12/3 = 4%.

Across all the non-d4 opening lines shown, adding a 2nd symmetrical move compared to the main line increases the draw rate 20/3 = 7%.

Although these statistics are about the same as in my previous thread, and therefore not overwhelmingly convincing, the pattern and values have at least remained relatively constant.

(27) an example of a temporal sequence: banrilotoq

banrilotoq = bishop attacking knight resulting in loss off tempo on queen. This is another weird name of mine that sounds like an Eskimo-Aleut word. This is another temporal pattern I noticed, especially in the Scandinavian Defense, but also in other openings. The reason I noticed it is that it seems odd to me for Black to capture BxN when that just forces his queen to get chased off when White recaptures with BxB, but that is the main line in some openings, so maybe there is logic I'm missing. I'm just presenting it "as is."

[Please allow 30 minutes after this repertoire post appears for me to notice and fix any typos and inaccurate wordings.]

Wow, you've gone incredibly deep with this. TBH, I don't think it's the best way to study or play chess for >99% of humans, but maybe this kinda stuff is going to be the foundation of the first computers to play based on "concepts".

Thanks for your comment. I feel like I'm forced into such studies because I can't progress rapidly with any existing materials I've found, especially regarding plans for openings, so I have to start nearly from scratch, which means looking at the very basics. It's like clearing a road before I can drive it at high speed. As for computers, yes I have an extremely heavy academic background and work background in computers, and although I didn't mention the connections, much of my notation and concepts are pulled directly from computer science, especially from the subfield of artificial intelligence. Does it show? 

Yeah, I think it does. Well, as long as it works for you, then keep it up!

To me your approach seems to be too formalistic. What is the advantage over simply doing an opening tree? The last two examples you give transpose to the same position after 8 Qe2, in the last one you mention six (!) names for the variation(s) and in the one before five (!). Most of these names IMO don't make sense like 1d4 Nf6 Gligoric defense or 1c4 e5 Timman defense, because they are not common names. This exhaustive naming of moves seems waste of space and work to me.

Regarding plans, wouldn't it be more fruitful to compile a database with model games for each opening you play and extract the plans from these games?

Regarding winning statistics, if you dive deeper into an opening tree where the number games decreases to about 10 you can find often that games were played between players with high rating difference or decided by blunders. So you should be careful relying on such general information...

I'm not so sure. Did you see the following post and quote a few days ago?

https://www.chess.com/forum/view/chess-openings/best-way-to-learn-openings--suggestions-please?page=5

"... In general, working by yourself with a database is probably more helpful than a poor opening book, but a good opening book can open up new vistas that you would probably not discover by yourself. ..." - GM John Nunn (2006)

That describes me well. I'm discovering a whole world in a single database, it seems. Of course I'm always looking at chess books, too, but not to the extent I hang out on move databases, often hours each day, looking for patterns that nobody else documents, and studying entire games in those databases to see how each story ended.

I'm going to be busy for three days again, so I might not get a chance to post in that time or to clean up errors in the above posts, like the errors involving transposition probems I discovered after I posted.