New insights in how to learn Tactics and better M2 recognition
So chess isn't about rules but in creating 'habits' to solve certain problems. I'm still in my 'checkmate in 2 phase'. For now I'm seeing most (80percent) in a instant.
So solving M2 is becoming quite boring. But besides being better in solving M2 i've learned some very valuable insights.
M2 isn't rocket science. But I still can't solve/see certain solutions. The reason is it that in some positions there are just too many candidate moves and I lack a good selection method.
Luckily are these problems not part of a normal chess game but composed puzzles. So why even try to solve ? Too complicated and they not really will improve my game.
But I also learned that after learning patterns Selection becomes a issue. If there are too many posibilities AND there is poor selection there is failure. Also if a pattern isn't that frequent in the DB there is failure.
Silent moves and frequency of position.
Well solving the obvious becomes obvious. Most M2 are just the same pattern. But I only have a few positions with M2 started with a 'silent move'. Still i'm seeing a backrank mate more often. So the best thing to do is to generate and focus on silent moves in M2 and to solve only this positions. Eventually it all comes down to selection and not calculation (anyone can do 4ply me guess). Think about this and the problem is reduced to a sorting problem !
How to learn Tactics.
It's still a long way until I'm on the level of tactics with the new method because after M2 i'll do M3 and then tactics (a year or more). But now I know for sure why people keep jojo ing in tactics AND i know how to learn tactics very effective and cummulative !!
It's crazy that after a 10 good solves the rating goes up and one has to solve more difficult problems. I already said that this is counterproductive. But now I know why ! Calculation is Shortterm memory and sometimes/somedays one is way better in that. Hence rating goes up. But nothing really changed. However if the selection process becomes better there really is a change. But this takes a lot of time. Eventually when new patterns are used the selection process becomes a issue. For a long time I didn't know how to solve this problem but now I know. And this means that learning tactics becomes less complicated !
Huffman decision tree and tactics.
I'm not saying that calculation isn't necessary but it isn't the main thing. Making a good selection is more important than calculation. With that in mind I started to understand that learning 'skewers / pins / vorks etc' helps solving tactics but isn't effective. So what is ?
A while ago I create a generalized notation of move sequences. which means so much as a relative abstract chessmove description system. So It's possible to look for instance for only tactics in which the first move is a queen move with check. (Q*+ in that notation). For some reason I got the idea to combine this with a huffman decision tree algorithm (with the difference that the student is the one that makes the decision) and came to the conclusion that this is the most effective way to learn tactics.
The process of 'learning' tactics
So the process to learn tactics is as follows.
1 Generate Tactics database from a small opening sequence with Stockfish vs Stockfish
2 Start by learning Tactics with certain movetypes only Selection=0
3 Let it become a habit , and gradually increase the movetypes like for instance (in GCNS=globalchessnotationsystem) first Q0+ and then Q0+ QN+ and then Q0 QN+ N0+.
4 iterate on that until all the movetypes are recognized.
Point 3 needs some explanation.
Selection works best based on 'habits=patterns learned until subconcious recognized' . With the old method the random tactics selection group is way to big to from 'habits'.
Well people already know that and that's why there is focus on thing like 'skewer' 'vork' etc.
But this method is way more precise by learning a same subset until this becomes a habit selection isn't stressed that much. (how many people don't even see a fork after 1000 tactics puzzles). There is a strong indication that stressing of selection works counterproductive , look I'm not saying that this method will not stress the 'habits' but it's less likely because the subgroups are way 'bigger'. For more info google 'pimsleur sequence PDF' (a alernative way of learning new items ).
Maybe defining a tactic sequence like that sound like cheating (and in a way it is in the beginning). But at the end of the learning it isn't , simply because the sheer amount of possible sequences that have to be learned (gradually)
The goal and how much time to 'learn' tactics
Eventually 'seeing' tactics up to 4 ply becomes possible in the shortest amount of time. In total there are for 4ply 3600 posibilities based on GCNS. Many of these (I guess half of it) aren't used there are 1800 or so combinations. To create a 'habit' it takes (if same speed as i'm learning m2) about 12 times for every possible position. With a average of 20 seconds per position it is...
1800x12x20 /3600 = 120 hours. with 1 hour a day it must be possible to 'learn' most basic tactics in 4 months..
Note : Only Basic tactics 4 ply deep , and not deeper
After M2 comes M3. As I guess right now M3 is about 5 times more complicated. So it will before I do tactics spend a lot of time doing M3 first before doing Tactics. At least 3 months.
I've learned valuable insights in doing an endless stream of M2. The real problem is it that most people want to advance too fast. The only way to advance fast in chess is to get the basics right first ,generate habits and augment the selection process. This is a very tedious process. Some people are lucky and can calculate very deep although calculation is important selection becomes even more important. The depth of calculation cannot be changed so to advance in chess , selection process / habits / pattern recognision must be enhanced.