Question about fast time control and losed games?

[Kempelen]

As I can not afford to buy new hardware, I will need to lower time controls in order to play more test games in a day. My question is, will 'losing on time' games be more counterproductive? I mean, Will I have more "noise" when calculating elo, that falsify the result? are there solutions?
Is this something usual or may I need to find appropiate opponents and prepare my engine to lose less games in time?
Fermin

[jdart]

Yes, losing on time will distort the results. And so if your engine is losing on time in short time control games, you need to fix that. (Same with opponents; you should not use those that have this problem).

--Jon

[bob]

As I can not afford to buy new hardware, I will need to lower time controls in order to play more test games in a day. My question is, will 'losing on time' games be more counterproductive? I mean, Will I have more "noise" when calculating elo, that falsify the result? are there solutions?
Is this something usual or may I need to find appropiate opponents and prepare my engine to lose less games in time?
Fermin

Losing on time games can skew the results. I found serious problems with this in my cluster testing, where some programs simply can not do fast time controls. And unfortunately, those are all losses for the program that screws up, making your program (assuming you are not losing on time) look stronger than it is. It has another effect of increasing randomness as well, as it won't always lose the same games on time, so sometimes it will lose losing positions, and sometimes it will lose winning positions...

For reference, in my 30K game tests, I generally see zero games lost on time, occasionally I might see one (There was a bug in stockfish that made it lose the same game every time, some sort of loop I presume, but I simply replaced that position to get rid of the problem).
\