sje wrote:If they wanted to clean up FICS, they would have done a better job with the same effort by booting every allegedly human account that had scored a victory over a program with twice the elo rating.
You might want to add the below formula for you engine, I have found that it gets rid of this type of abuse.
Bryan Hofmann wrote:You might want to add the below formula for you engine, I have found that it gets rid of this type of abuse.
rated & assesswin > 0 | unrated
Done. Let's see if it helps.
--------
Had one FICS game earlier today with opponent some 1000+ elo less than the program, but he somehow managed to do well -- for a while. But for some reason his quality of play took a big drop and he found himself facing a certain mate in one. First, he asks for an adjournment. Then he asks for an abort. After that he offers a draw! Finally he disconnects.
For some reason, the "assesswin > 0" in the program's formula is not working. The program is still being matched to very low rated challengers. I tried "assesswin > 1" and there was no difference.
--------
One idea is to equip Symbolic with a cheater detector of some kind. For example, if a supposedly human ICS opponent performs at, say, 1200+ elo above his/her standard rating, then the opponent gets automatically added to the noplay list.
--------
The FICS handle "crafty" hasn't logged in for over three years. Its finger notes say that it's running on a 1 MHz z80 with 64 KB RAM.
sje wrote:For some reason, the "assesswin > 0" in the program's formula is not working. The program is still being matched to very low rated challengers. I tried "assesswin > 1" and there was no difference.
--------
One idea is to equip Symbolic with a cheater detector of some kind. For example, if a supposedly human ICS opponent performs at, say, 1200+ elo above his/her standard rating, then the opponent gets automatically added to the noplay list.
--------
The FICS handle "crafty" hasn't logged in for over three years. Its finger notes say that it's running on a 1 MHz z80 with 64 KB RAM.
The z80 reference was a joke by Bob, mocking some accounts that were under-reporting their hardware at the time.
sje wrote:For some reason, the "assesswin > 0" in the program's formula is not working. The program is still being matched to very low rated challengers. I tried "assesswin > 1" and there was no difference.
--------
One idea is to equip Symbolic with a cheater detector of some kind. For example, if a supposedly human ICS opponent performs at, say, 1200+ elo above his/her standard rating, then the opponent gets automatically added to the noplay list.
--------
The FICS handle "crafty" hasn't logged in for over three years. Its finger notes say that it's running on a 1 MHz z80 with 64 KB RAM.
Make sure you seek ads for games contain the "formula" or "f" variable.
Bryan Hofmann wrote:Make sure you seek ads for games contain the "formula" or "f" variable.
Ah, I had forgotten that bit of minor arcana. I've become acclimated with the use of the "useformula" variable on ICC.
Alas, with the new seeks the program is only seeing candidate counts in the low single digits. I'll have to do more tuning and perhaps add the cheater detector.