Fritzlein
Forum Guru
Arimaa player #706
Gender: 
Posts: 5928
|
 |
Movement of rating points
« on: Aug 5th, 2005, 12:59am » |
 Quote  Modify
|
99of9 has pointed out that, since the large majority of games are human vs. bot and only a few are human vs. human, the transfer of rating points from one human to another occurs mostly via the bots. It is as if the bots are a big bank of rating points, and some people are depositing points while other people are withdrawing.
To put some numbers to this thoery, I chose ten recently active players, and for each measured their rating change over their last 100 rated games. I divided their games based on opponent type, and compared the net change of rating in games against bots to the net change of rating in games against humans. Only one of the ten players (robinson) had a greater magnitude of rating change from playing humans, so the bank account theory holds at least that far. We mostly trade points via bots rather than directly trading with each other.
Here are the changes sorted by difference in performance versus bots as opposed to versus humans:
| Player | Rat Chng | vs. Bots | vs Humans | | omar | -287 | -226 | -61 | | naveed | -19 | -62 | +43 | | Belbo | -41 | -68 | +27 | | robinson | +90 | +32 | +58 | | 99of9 | -98 | -56 | -42 | | blue22 | +63 | +34 | +29 | | Fritzlein | +57 | +56 | +1 | | PMertens | +54 | +77 | -23 | | haizhi | +78 | +95 | -17 | | Ariminator | +400 | +413 | -13 |
This table seems fairly intuitive to me: the five of us in the bottom half did a lot more bot-bashing for rating points than the five in the upper half.
By the way, the my data is a couple of weeks behind, so it may not include some salient recent results, if folks are wondering why it doesn't exactly match their memory of the most recent rated games.
|
Home
Help
Search
Members
Login
Register
Reply
Notify of replies
Send Topic
Print
Author
IP Logged