You scored around 50% against the 1200s, so maybe 1200. p.s. I think you should play against 1350-1400 rated guys.
Statistical Analysis of a Bullet Player
Sort:
Well, I scored 46.5% against all players rated 1201-1250, so that would imply my break even mark is probably 1230.
Yeah, looking at these stats I do just as well against 1351-1400 players as 1251-1300 players. My theory is more players between 1351-1400 bolster their ratings in the method I discussed than 1251-1300 players, so overall they're probably about the same.
I'll test this theory too!
Can't say. I only played bullet twice and since I won both games while having a really high rating-uncertaintanty, this resulted in my bullet-rating being my best rating.
So far I have 100% win-ratio against everyone!
So not playing it anymore seems reasonable to keep my perfect-stats. 8[
DJDopamine wrote:
When you play these games, do you often play a particular opponent multiple times? My thinking is that a player may have managed to obtain a rating that doesn't match their ability. Playing against someone like this repeatedly will affect the data. You can also "learn" how they like to play (opening choices, etc.) and improve your performance. Lastly, some players seem to do better against certain "styles" of play, and therefore will do better against some players. My other question is: How was each game won (Checkmate or Time), and what the distribution of this would look like. It may be reasonable to expect more "Time" losses at lower ratings, but I could certainly be wrong.
Whenever I beat an opponent rated higher than myself I'll always challenge them to a rematch immediately, as that makes sense from a ratings perspective. When I lose to a lower rated player I won't, though I'll usually accept a rematch if they offer it.
I've probably scored 10+ easy wins against n00b 1200-players, who are new to Bullet, and lose while I've still got 50 sec on my clock.
What I find with 1-minute Bullet games is opponents between 1200-1500 are often relatively similar in strength, but the higher rated ones tend to manage their time marginally better, and therefore win in the end. I would analyse this, though I don't think there's any way to extract time vs. checkmate wins/losses except by visiting each game separetely.
I've probably lost 1000 games on time, mostly to higher rated opponents.
Yesterday, I lost this game, frustratingly, they had only 0.7 sec on their clock: http://www.chess.com/livechess/game?id=771886187
As for openings, I play the Colle Zukertort as white, as I can play the same moves very quickly, and it sets a nice trap for black. The speed I play it also gives me a chance to pressure an opponent on time later in the game. http://www.chess.com/article/view/quick-start-guide-to-the-colle-zukertort-an-opening-for-players-of-any-strength
As black I don't have much of a strategy, except deploying my pieces somewhat unconventinally. I tend to castle queenside a bit as well, which can slow some players who have deployed kingside.
Perhaps my biggest failure is a tendency to premove too much, especially when a black player takes a few seconds to start moving. On a few occasions a black pawn has ended up on the back rank while I'm sitting there watching my premoves ruin my game!
WISH_I_WAS_A_GM wrote:
You know we played once PDubya,
Not my proudest moment. I wish it recorded the time remaining after each move, as that would be quite useful when looking back at these moves.
After some experimentation, I conclude that playing against 1100-1700 is the best way to get a good indication of my rating, which is currently at 1277.
I like your approach, would you add me and perhaps settle for bullet ?
punctate wrote:
I like your approach, would you add me and perhaps settle for bullet ?
Unfortunately, I don't think it would be much competition for you, given you're currently at 1737. I may draw or win on time once in every 50 games. Probably not good for my rating!! But we can always give it a go...
After struggling for a long time to improve my rating in Bullet (dropped to 1203), I decided to change computers, and played on a fast desktop PC at my workplace (after work). Prior to this, I played on a Mac with a touchpad mouse, or on my phone (no premoves). The results were immediate.
Starting at rating 1203, I played players rated 1300-1500 (this averaged out to around 1350) and scored 15/24, including 9/13 with White. This was a huge turnaround, and it became very obvious how much equipment influenced this type of chess.
Based on this, I can expect my rating to converge somewhere around 1430 if I keep up this level of play.
Has anyone else seen a sudden increase in rating after changing PC?
I'd imagine a good mouse and a fast connection (compared to the alternative) is easily worth 100-200 points.
Back on my MacBook with a crappy touch pad and I drop the 100 points I made, struggling (and losing) to several 1000 rated players. Damnit!
Buy yourself a logitech g400s gaming mouse, PDubya. That will be worth at least 100 points.
Also solve at least 10 tactics a day. That will also be worth at least 100 points.
Finally, when you want to practice, challenge Computer medium or hard to 1|0 unrated. Keep trying to improve the score against the engines. I used to go 2/5 against Computer medium but now i can usually go 5/5. It helped me tremendously when I practiced 1min with the engines.
Thanks for the tips, GM. I'll definitely do that! I should also avoid playing late at night, or when I'm in a bad mood. I went from 1350 (should have stopped playing here) yesterday to 1264 because I kept saying just one more...
Can confirm. Good computer mouse is a lot. Still waiting for it. If only I could move the pieces with my mind. :)
The rating system used by chess.com is statistical itself and, as soon as one plays enough games to measure the performance more or less precisely, they have a ~50% winrate vs equally rated players *by definition*.
Read about the simpler Elo system http://en.wikipedia.org/wiki/Elo_rating_system first to get the idea, it's based on the assumption that a player who's by 200 points stronger than me will have a ~75% winrate vs me. The Glicko system http://en.wikipedia.org/wiki/Glicko_rating_system - the one used by chess.com - additionally takes into account the inaccuracy of rating measurement that slowly disappears after a big number of games.
Bullet players surely get more accurate personal stats faster as they play more games a week, but this distribution of winrates (I mean % of wins + half of % of draws) is believed to hold for other speeds and variants too.
My "true" rating is probably around 1200, and I have found that the best way to increase my rating is to play against stronger players, such as players in the 1250-1400 range. I don't win as many games but I usually see a rating increase.
My rating tends to stabilise around the 1300 mark now. I'd like to do some mass number crunching again, but it's hard to get this, as we can only copy and paste 50 games at a time. Very restrictive!
H-A-N-D wrote:
My "true" rating is probably around 1200, and I have found that the best way to increase my rating is to play against stronger players, such as players in the 1250-1400 range. I don't win as many games but I usually see a rating increase.
The rating is artificially increased by playing with overrated opponents, not necessarily strong. An opponent rated 1250-1400 (only a bit higher than the starting 1200) might be a weaker beginner who has accidentally won a few games, resulting in a higher rating than the one corresponding to the skill level.
That said, facing stronger players certainly helps our own true strength grow too, as they force us to think better.
I'm a relatively frequent Bullet player, and my rating hovers around the 1250 mark.
http://www.chess.com/livechess/stats/PDubya?type=Bullet
I wanted to know how I performed against players of different ratings, but that wasn't readily available in the Stats area, so I broke it down myself. The results were interesting.
Opponent Games % Score
1001-1050 13 100.0%
1051-1100 29 93.1%
1101-1150 79 77.2%
1151-1200 103 54.9%
1201-1250 113 46.5%
1251-1300 81 35.2%
1301-1350 36 33.3%
1351-1400 33 33.3%
1401-1450 8 12.5%
Against sub-1100's I almost never lost.
Against 1101-1150's I was coming in at more than 75%.
Once I hit 1151-1250 players I came out about 50-50, and then it just descended into really bad numbers from there.
Colour didn't really matter, though I actually found my overall performance was slightly better with black.
I think most players would have a similar pattern, and it's at the point where we're scoring 50-50 that we find our true rating.
If I wanted to beat up on patzers under 1100 all the time I could probably get my rating up to over 1400 (I may try this in the interests of this post).
If I only played over 1400 players I'd probably be down around the 1000 mark (I'm less inclined to do this, though it may actually make me a stronger player).
Most of the time it's possible to feel how good a player is, rather than notice their rating. I've had a few wins against 1500+ players, and I am fairly sure they built their ratings up artifically.
So, based on this methodology, what's your true rating?