Sudden Death - Improve TM

[TierynnB]

Due to the 50 estimated move horizon, once a sudden death game got below 1 second, the time allocation for optimumTime would go into the negative and SF would start instamoving.

to counter this, once limits.time is below 1 second, the move horizon will start decreasing, at a rate of 1 move per 20ms. This was just what seemed a reasonable rate of decay.

Fishtest sudden death TC 5+0
https://tests.stockfishchess.org/tests/live_elo/65ee2cdf0ec64f0526c43bbb
LLR: 2.99 (-2.94,2.94) <0.00,2.00>
Total: 3348 W: 1070 L: 727 D:1551
Ptnml(0-2): 46, 277, 738, 514, 99

Fishtest SD TC 10+0
https://tests.stockfishchess.org/tests/live_elo/65ee401e0ec64f0526c43cf7
LLR: 2.95 (-2.94,2.94) <0.00,2.00>
Total: 3780 W: 1097 L: 808 D: 1875
Ptnml(0-2): 11, 353, 919, 550, 57

Neutral Non-Regression STC 10+0.1
https://tests.stockfishchess.org/tests/live_elo/65ee45ff0ec64f0526c43d68
LLR: 2.95 (-2.94,2.94) <-1.75, 0.25>
Total: 123616 W: 32054 L: 31927 D:59635
Ptnml(0-2): 493, 14323, 32105, 14338, 549

neutral Non-Regression LTC 60+0.6
https://tests.stockfishchess.org/tests/live_elo/65ef1eec0ec64f0526c44bc4
LLR: 2.95 (-2.94,2.94) <-1.75, 0.25>
Total: 130482 W: 32961 L: 32855 D:64666
Ptnml(0-2): 88, 13412, 38123, 13542, 76

[TierynnB]

For reference, heres an example of how the optimum/maxtime were before and after this change at times below a second.

		Base code		Change Code
inc(ms)	limit.time(ms)	OptTime(ms)	MaxTime(ms)	OptTime(ms)	MaxTime(ms)
0	1000	20	128	20	128
0	900	16	99	18	113
0	800	12	70	16	98
0	700	7	41	14	84
0	600	3	12	12	69
0	500	-9	0	10	54
0	400	-9	0	7	34
0	300	-9	0	5	21
0	200	-9	0	3	9
0	100	-9	0	1	-2

[Disservin]

did we have patch on fishtest which scales the mtg accordingly ?

[FauziAkram]

did we have patch on fishtest which scales the mtg accordingly ?

@Disservin yes, I tried something like that long time ago, let me find the links.

[FauziAkram]

https://tests.stockfishchess.org/tests/view/65286c5b3125598fc7ebaaaf
https://tests.stockfishchess.org/tests/view/65295c533125598fc7ebbe3e

[TierynnB]

I had tried something similar, just a linear curve from 50 to 30.
https://tests.stockfishchess.org/tests/view/65d5266e1d8e83c78bfdc4d4
https://tests.stockfishchess.org/tests/view/65d9d1af4c0e22b904f5752b

I think maybe from this change, increase the time it starts to degrade(and adjust the rate), and see if it has a positive impact.
Idk how big an impact it would have tbh.

[Disservin]

https://tests.stockfishchess.org/tests/view/65286c5b3125598fc7ebaaaf https://tests.stockfishchess.org/tests/view/65295c533125598fc7ebbe3e

what are these tests :o
I was imagining something like @TierynnB later posted. Looks interesting and is perhaps the correct thing to do. I'll merge for sudden death, but it be nice to replace this in the future with a more general approach.