### Limiting nodes to reduce engine strength

Posted:

**Sat Oct 26, 2019 3:22 pm**I'm trying to use the method explained by Ferdinand Mosca here :

viewtopic.php?f=2&t=71199

Limiting nodes and running 1m+1s games based on Cheese 2.1 = 2811 ELO CCRL :

Using logarithmic regression :

maximum nodes = e ^ ((elo + 838.8755655) / 242.5737137)

But when I run games with other engines I have different results :

What engine to use as base to calculate ELO ratings ? or I have to find the rating of each test version one by one ?

The curve from the equation is not perfect compared to the game results, there is a better equation, or should I do a linear interpolation between result at different ELO ratings (ex: 1500->1600, then 1600->1700 ...) ?

viewtopic.php?f=2&t=71199

Limiting nodes and running 1m+1s games based on Cheese 2.1 = 2811 ELO CCRL :

Code: Select all

```
engine nodes elo elo from formula
Cheese-22-64-test1 500 472 668
Cheese-22-64-test2 1000 693 836
Cheese-22-64-test3 2000 958 1004
Cheese-22-64-test4 4000 1193 1173
Cheese-22-64-test5 8000 1426 1341
Cheese-22-64-test6 16000 1611 1509
Cheese-22-64-test7 32000 1798 1677
Cheese-22-64-test8 64000 1970 1845
Cheese-22-64-test9 128000 2132 2013
Cheese-22-64-test10 256000 2288 2181
Cheese-22-64-test11 512000 2419 2350
Cheese-22-64-test12 1024000 2536 2518
Cheese-22-64-test13 2048000 2639 2686
Cheese-22-64-test14 4096000 2746 2854
Cheese-22-64-test15 8192000 2803 3022
```

maximum nodes = e ^ ((elo + 838.8755655) / 242.5737137)

But when I run games with other engines I have different results :

Code: Select all

```
Rank Name Elo + - games score oppo. draws
1 Cheese-22-64-test5 1618 16 16 400 66% 1501 18%
2 Piranha 0.5 1501 16 16 400 35% 1618 18%
Rank Name Elo + - games score oppo. draws
1 CDrill 1800 1786 16 16 400 55% 1747 15%
2 Cheese-22-64-test6 1747 16 16 400 45% 1786 15%
```

The curve from the equation is not perfect compared to the game results, there is a better equation, or should I do a linear interpolation between result at different ELO ratings (ex: 1500->1600, then 1600->1700 ...) ?