in what might very well be one of the most suicidal moves in computer chess history, I'm asking for help to make my program weaker!
Yes! I'm not crazy, no more than usual at least... read on! 
Well it is quite simple really. The constant race to get more Elo and strength is nice and all but at times I really miss being able to play and engine and get some fun out of it.
Some engines do support a "limit Elo" feature but that kind of play is not always satisfying. Sometimes, you realize you have blundered badly... engine's value goes up to +5.0 then it plays a dumb move and restores the score close to zero, makes you feel really bad!
This beta version of Hamsters is an attempt to write an engine that plays weak, but not in an exceedingly obvious way. There are several shortcomings in this version, so it might play a bit dumber than I would like. I hope everything will be fixed before the final version though.
Now this beta is for everyone that wants to try this feature and send feedback, anything from style to Elo strength and so on. You need an UCI compliant GUI to play, as there are several parameters to tinker with. Many of them will disappear in the final release though, hopefully replaced by predefined personalities and/or the "limit Elo" parameters. Here's a list:
Limit KNPS. Limits the number of nodes per second searched by the program. This is useful to keep everything "objective" and independent on hardware specifications, as long as there is enough power to run at the specified NPS. Note the parameter specifies Kilo NPS, so a value of 100 (the default) will have the program search 100,000 nodes per second. I think this is an honest way to limit strength, better than limit nodes or depth. Anyway, Hamsters will of course honor also those settings if specified (elsewhere).
Pick Percent Min and Pick Percent Max. Those two parameters control how good is the move that the engine chooses to play. After each iteration all available moves are sorted according to their strength, with the best moves at index 0 (i.e. percent zero) and the rest of the moves following. Setting Pick Percent Min to 0 and Pick Percent Max to 10 for example means that the engine will pick a move from those that fall between 0% and 10%, that is a random move between the best 10% available. It is possible to change the interval range as well as shift the entire interval. For example setting Pick Percent Min to 10 and Pick Percent Max to 30 will force the engine NOT to play the best move, but rather some inferior choices.
Note however that there are compensation mechanisms in play, that will override the above rule if deemed necessary. For example, the engine will try to avoid a downright blunder caused by the above rule.
Blunder. This parameter tells the engine to play a blunder every now and then. A blunder does NOT mean it will leave its queen hanging, you'll have to earn it! In other words, the engine will try to setup a mini-combination for you to find, and this parameter controls how deep the combination is. Of course it's not guaranteed that the engine will always find such a combination, or that it will blunder every time it sees one. However, it will almost always blunder if it thinks the blunder happens at a more difficult level than you have specified, just to challenge you!
Hide Score and Hide PV. Those hide the score and PV of the engine, which is exactly what happens when you play a human. This behavior will be controlled automatically in the final release, but the controls are here in the beta for debugging purposes.
A ton other settings... control the engine evaluation. They are here just to play with and see if some "personality" can be more fun than the default, but until now I've done very little testing on this.
In fact, the whole thing is not fully tested yet. Because all of the test must be performed manually I'm going quite slowly. Yet, it probably works well enough to be fun... at least I hope it is!
Beta version is available for Windows only:
http://www.ascotti.org/programming/ches ... 3beta3.zip