Reinventing the wheel: Chessmaster type personalities on Chess.com

What are your thoughts upon the introduction of Bot personalities in the Play Computer area of the site?

Immediately, I was reminded of the Chessmaster series and also Kasparov Chess for the Gameboy Advance that used a similar personality level system.

Does the convenience of a pre-tweaked engine make it more likely you’d play a personality system? Do/did the back stories hold any relevance to you.

I think the key questions have to be:

1) Do any of the Chess.com personalities remind you of someone you play?

2) Are these new personalities an improvement on the old Chessmaster personalities?

Funny enough, I responded to a chess.com forum post regarding chess personalities a month before today's announcement of this new feature here in chess.com. Coincidence?

Don’t know. I learned about the feature in the Chess Community forum via a thread started by DeepGreene. I haven’t updated the app for a long while so the feature was not visible to me.

I think it is a good addition even though computer personalities are not a new idea. Been playing each personality open to me until I get stuck. Then I can try ChessMaster DS (as the PC version won’t work on Windows 10) and compare results.

A number of engines have had personalities since forever. The question to me is not whether they are better than Chessmaster, but whether they are better than what exists, or has existed, here!

HIARCS is supposed to be more human-like in terms of play. On the app, you can select between 3 playing styles (aggressive, active, and solid) and 5 levels of book knowledge (off, wild, surprise, dynamic, and tournament) for a total of 15 types of opponents. And you can select a rating from 800 to 3200 in 25 point increments.

I think it would be great if all of that could be done on this site (and be able to save and analyze those games on the site). But there is no doubt that HIARCS is superior in terms of mimicking human play at the lower levels ... the computer on this site does things like hang a queen, then starts playing like a super-GM for several moves, then back to a blunder that no 800 player would make. At 800-900, I have had HIARCS drop an exchange, but in the middlegame, not on move 5 just planting the queen on your 3rd rank in front of a row of your pawns. Ridiculous.

I have been working through the new personality bots on Chess.com and played a couple of games on Chessmaster for the Nintendo DS as that is the only working version I have.

All games were untimed.

I won against Nelson (1300) but lost against David (1400). This would put my own score at 1320 ish as Nelson was a good adversary for me. Both bots are interestingly positioned in the “intermediate” category when their estimated ELO scores would classify them as Novices.

On Chessmaster DS however, while I did win after a massive 88 moves, I felt well matched against Rand (1007 est ELO)

Meanwhile, Shredder, the mobile app, has an estimated rating for me at 1275 after 75 games, 1272 after the last 10 games.

While I like the personality bots on Chess.com I do wonder if Chess.com is chasing the endorphin rush classifying their novice bots as intermediate and setting their “estimated elo” slightly above their actual ability?

Chessmaster DS does take a surprising amount of time to compute moves at the lowly rating of 1007 and there is no option to make it move instantly. The wait does not bother me as I am usually thinking about the position. While the bots have wait timers instilled, they still move too quickly for the ratings listed above. However, if realistic wait times were programmed in, it should definitely be an option that can be turned off for those who do not want it.

As for feel of play, personally, I think the personality bots on Chess.com are an improvement over Chessmaster DS in terms of how blunders and mistakes are incorporated.

All that chess.com needs to do now is to rename the corresponding low elo strengths to the up-and-coming versions of prominent players akin to Young Fischer, Pre-pubescent Kasparov and Awkward Carlsen.  

 

I'm working through the bots as well, currently at Pablo :-)

My impression is that they are trickier during the opening / middle game,

and get quite a bit weaker in the endgame, do other people see this too?

 

It may just be my weak opening repertoire, etc. though.

 

 

quite

Just did the same thing with Pablo (who's supposedly 1600 ?!): clean up the board quickly,

he makes a silly mistake near the end so it's rook against bishop, and then he doesn't defend his pawns properly.. (he had 4 pawns against 2, I could just walk over with the king and clear them)

 

I have downloaded the Hiarcs chess app and while there are many useful features in the app, I am unable to find the 15 opponents you speak of other than the Hiarcs chess engine.

Could you explain how I get to them in the menu system?

Thanks.

They're not pre-defined opponents per-se - what I think dannyhume is saying is that there are 15 combinations of playing styles.

To configure these in the Hiarcs app, select New Game and then tap "Hiarcs" at the bottom, then "Engine room". Once there, you can set the Elo strength and also "Playing Style" ("Solid", "Active" or "Aggressive") and the "Book Variety" ( "Off", "Wild", "Surprise", "Dynamic" or "Tournament").

With 3 "playing styles" and 5 "book varieties" you have 15 combinations of these.

Do you think the higher rated personalities have more accurate ratings? I have played all the personalities up to and including Charles...who I have now played a few games against. I only move on to the next personality once I have beaten a particular one. I would like to think that this will help me as a tool in my improvement, hope its not there just for vanity. 

Ho, OK, understood. I thought he meant they were like Chess.com or Chessmaster. It is still a feature rich app though so I am pleased it was recommended.

My Saitek Chess Academy has 60 fun levels, each level getting progressively harder. In reality each level feels different from the last in that a different aspect of strategy is tested. I suppose one could think of each level as a “personality” without the mug shots and short descriptive passages that may or may not match the personality’s playing style.

In fact the manual gives away no clues as to the specifics of the fun levels or their attributes.

I think the fact the bots on chess.com play less like a chess engine and at lower levels provide those “ho, okay” moments that weaker players provide by hanging pieces and not defending weak squares puts them ahead in my book.

Given I am having trouble beating Maria even though I have beaten her neighbours does make me think the allocation of ELO levels is either rushed or too theoretical and not properly tested.

I think it is the uneven ELO allocation that is causing most of the head scratching amongst commentators. At least that is to say that is the aspect of the personality bots I see most discussed.

The chess.com bots - including Maria-BOT - are playing rated games, so you can see their real strength against the chess.com rating pool. Currently Maria-BOT is Blitz 977, Rapid 1015 and Bullet 1261

So if I am playing the bot via Computer and untimed, does the stated ELO not apply in that situation? One would assume the slower the time the stronger the bot?

I am a bit confused why a bot on a server that should amount to a super computer and therefore able to compute its move in a millisecond would play at different strengths at different time controls? Surely the ELO score of the bot in rated games is more dependant on the pool of players it finds itself pitted against?

Anyway, even with that variable I am interested to compare Maria’s neighbouring bots to see if my perception that Maria is playing a harder game than those bots in her immediate group is true.

I don’t use the main site very often so could you explain how I view various bot ratings as you did for their rating pool?

I'm just waiting for someone to unlock the Hikaru-Nakamura-Bot.