Correlation between Tactics Rating and Daily Rating

I've got way too many degrees, and almost all of them required that I take statistics.  Even when I started in a doctoral program I had to take statistics again.  So there are times when I look at numbers and questions crop up that can only be answered through a statistical analysis.

What is the correlation between players on Chess.com's Tactics rating and their Daily rating?  How strong is the correlation?

If no one knows, do you think Chess.com would be willing to supply me with an Excel spreadsheet of all the active players' Daily rating and their Tactics rating?  (I wouldn't need names or other identifying information, just those two columns.)

Actually, give the limitations of Excel, I'd only want 1,048,575 players' information.  After that, it won't fit on an Excel spreadsheet.  I don't own SPSS, and haven't installed R, which would be needed for a bigger data set.

Getting the data aside, just common sense dictates that there is a strong correlation between chess rating and tactics rating (or lessons rating or anything chess related for that matter). Of course, a better investigation would be revealing the fact that tactics ratings stronger correlate with bullet/blitz rating versus daily rating. Fast players are clearly tactically alert, but many observers often miss how positionally-talented many of these players are too.

I suspect that chess.com already has the data compiled you seek somewhere (maybe not an Excel sheet though). My question is: why do you seek these statistics? Do you intend you use these statistics to assist your chess, or is the reason purely for mathematical fun?

If it is to improve your chess, then HOW strong the correlation is hardly matters; it helps your chess, so a chess player would do it to improve. 

If it is just for the utility you receive from the math, then the chess related subject would have little importance anyway. Any math (especially statistics it sounds like) would be  satisfying to you; therefore, if you can't get the information you seek - then you will gladly simply switch to another topic.

From a practical perspective, I don't know if the effort and work is worth the benefit. If such statistics exist, I would imagine it was gathered by some computer program hooked up to a database.

If no engrossing math talk results from this forum, you could always try contacting chess.com directly. 

I know the correlation between tactics and daily ratings would be strong, but keep in mind that many strong chess players reset their tactics rating, or let their chess students use their profile for tactics. This may weaken the "true" correlation.

Nevertheless, good luck with your statistical adventures

Why?  Just curious.

I am virtually certain many chess databases already have this correlation, and even more certain that the correlation is positive and strong. How strong? I am not a database XD but I also don't find the relevance in knowing this; and this comes from a philosophy lover where relevance is often just a bonus!

I mentioned this (buried in my monsterously long post - sorry everyone) earlier.

Statistically though, a correlation of some kind would still exist - even if ALL tactics ratings we're reset. Of course, the correlation would not accurately reflect the true correlation though. Hypothetical scenarios aside, I am pretty sure that the true correlation on chess.com is strong - even with the accounts that reset their tactics rating.

Dude , if you would like to do statistical correlation, you better do conditions with equal standarization among participants.

In daily games, some people use 30 sec per move, some people use 30 minutes per move. (more variable conditions among participants). And also there are a high percentage of unfinished games with time out, close accounts etc.

 

You should rather use blitz game where participants have the same standarization, less percentage of unfinished games etc  . ( equal time usage in blitz games, e.g)

There's a weak correlation because of certain tactic exploits that chess.com doesn't ban for, and daily has some variance too due to some people playing it literally like blitz, and some people take a long time on each position.

 

If chess.com released raw data that would be a lot of fun. Ask them and see what they say.

I was scrolling through looking for a comment like this.

I.e. people who don't know what the word correlation means saying there is none. Not that I'm trying to pick on you, I'm sure you wont be the last.

Ok, I agree that's a bit silly.

But let's compare.

Suppose someone said that even if you compare the sales of 1000 dealerships, there's no way to tell which sold more or less black cars based on their overall sales.

That's even more silly. Sure if what you're really interested in is sales of black cares there are better ways to do it, but overall numbers are going to give a good rough estimate.

Same thing for chess.com players. The predictions we make for individuals may be off by some 100s of points, but overall we can get a rough idea.

And BTW when I read this I'm thinking to myself Jesus, then why would you say there's no correlation? You want the OP to conform to your view of which ratings should and shouldn't be respected?

I don't take tactics ratings seriously either, but as an academic question it would still be interesting to get some raw data from chess.com and see what sorts of correlations we can make between rating types.

Ok, but if the top ratings are using exploits (and I'm sure they are) while >95% of players are using TT honestly, then you'll still be able to use the data to make reliable predictions.

You can even throw out outliers like people with a 4000 TT rating (I don't know what the highest ratings are, just making the point that the leader board is only a small part of the story).

When a tactics rating is reset does it also reset the number of tried/failed puzzles? If so, then that's not a big problem. Just filter the data for players who have completed many puzzles and are currently active in both TT and daily chess.

Your point #2 would screw it up though, you're right.

And in general, I regard TT ratings as the most unreliable, so I'm willing to accept that we'd end up with a bunch of garbage, as you say.

That's true, some people do so many tactics that they start to memorize a large portion of puzzles. So of course their TT ratings should be thrown out too. I guess you'd have to filter it from both ends. You wouldn't want anyone new, but also wouldn't want the most experienced solvers either.

I assume (you're right about that) that it's like a bell curve with ~85% of people in the middle of it. I don't imagine that we'd be throwing out so many that the sample would become too small.

Stats class was years ago, but from what I recall a sample size of just 1000 players would be more than enough... yes we're ignoring the extremes, but I assume people like the OP aren't so interested in e.g. total beginners who fail almost every puzzle, or addicts who have memorized most of what they're going to see.

Oh, you're right, I didn't realize the distribution is so absurd.

I see the OP is under 1000 in tactics as well.

You made good points, I agree that anything close to a random sample would be garbage.

By the way, @ the OP and anyone with a tactics rating like that, I don't think chess.com's tactics trainer is a good way to start. Maybe I'm old school, but I think it's better to start with a book that will explain a little bit about common tactical themes (pin, fork, discovered attack, removing the defender) and give you puzzles that have an actual theme instead of non-puzzles that are just capture a hanging piece.

omg

My final statistics classes were in a doctoral program for Public Affairs (specifically Public Administration).  The program was a social science program, as opposed to a "hard" science program.  In it, we were looking at correlations with R-Squared of significantly less than .5 as being significant, assuming the p-value is under 0.05.   That's the real test if there's a correlation.

If the R-Squared is low, that would imply that there are other factors involved, not that there wasn't a correlation between the two.  A correlation can be weak, but still exist.  If you did a full multi-variant analysis, you'd probably find several different factors that work together.  But I don't have a clue where you'd even get the data.  (Date joined Chess.com as evidence of years playing chess?  Nah, way too many other variables to use that as a proxy.) 

On a different note, it's interesting to see how quickly the conversation turned to my ratings.  I'm not sure what my ratings have to do with anything; I was just staring at my stats and wondering how strong the correlation is between the two that I'm working with.

I would say it would be better to correlate tactics training with an actual OTB rating.

You don’t know how much a given player gives serious thought to an online chess game for a given move. You don’t know who does and who doesn’t use the analyze board regularly. You don’t know who glances at her/his mobile device and makes a move based on a few minutes look at the position as opposed to who sits there thinks about their position, returns to it hours later, etc. You don’t know who uses their memory/knowledge for the opening versus who uses an opening book to whip out their first 8-12 moves.

Personally, I’d be interested in a correlation between tactics rating, ELO (OTB), and the types of tactical mistakes made, especially ply, tactics problem rating, and then motif(s).

I will use myself as an example.  I was an 1800 OTB player.  My tactics rating here is 2714.  I am NOT a 2714 tactical player.  In the software CT-Art 3.0 my "FIDE" tactics rating is over 2800. 

The only true measure of someones chess ability is OTB play at long time controls. 

hi i am new to the game