Difference between revisions of "User:Voidious/Robocode Version Tests"

From Robowiki
Jump to navigation Jump to search
(PrairieWolf alpha4, 500 more Raiko vs Moebius)
(→‎Alphas 8 & 9: Raiko vs Moebius, 1000 battles each)
 
(5 intermediate revisions by the same user not shown)
Line 4: Line 4:
  
 
The score % is for the first bot listed.
 
The score % is for the first bot listed.
 +
 +
----
 +
 +
=== Alphas 2, 3, 4 ===
 +
 +
* Alpha 2 uses the old decel through zero rules (can go from -1 to 1 or -0.1 to 1.9).
 +
* Alpha 3 uses the proposed new rules (splits the tick into decel and accel portions, eg -1 to 0.5 or -0.5 to 0.75).
 +
* Alpha 4 caps maximum velocity at 1 when decel through zero (-1 to 1, -0.5 to 1, -0.1 to 1).
 +
* Turns out all these Alphas contained an inconsistency compared to 1.6.1.4 rules, so they were scrapped for more Alphas and more testing.
  
 
<!--
 
<!--
Line 126: Line 135:
 
|align="right"| 60.18
 
|align="right"| 60.18
 
|align="right"| 60.03
 
|align="right"| 60.03
|align="right" style="background-color: #ccc; color: black"| '''0.28'''
+
|align="right" style="background-color: #ccc; color: black"| '''+0.28'''
|align="right" style="background-color: #ccc; color: black"| '''0.26'''
+
|align="right" style="background-color: #ccc; color: black"| '''+0.26'''
 
|align="right" style="background-color: #ccc; color: black"| '''+0.11'''
 
|align="right" style="background-color: #ccc; color: black"| '''+0.11'''
 
|align="right" style="background-color: #cfc; color: green"| '''-0.02'''
 
|align="right" style="background-color: #cfc; color: green"| '''-0.02'''
Line 175: Line 184:
 
|}
 
|}
  
 +
 +
----
 +
=== Alphas 8 & 9 ===
 +
 +
* Alpha 8 uses the old decel through zero rules (can go from -1 to 1 or -0.1 to 1.9).
 +
* Alpha 9 uses the proposed new rules (splits the tick into decel and accel portions, eg -1 to 0.5 or -0.5 to 0.75).
 +
 +
{| border="1" cellpadding="3" style="border-collapse: collapse; color: black"
 +
|
 +
| '''1.6.1.4'''
 +
| '''Alpha8'''
 +
| '''Alpha9'''
 +
| '''1.6 & A8'''
 +
| '''1.6 & A9'''
 +
| '''A8 & A9'''
 +
| '''Battles'''
 +
| '''Comments'''
 +
|-
 +
| ''Diamond vs Komarious''
 +
|align="right"| 70.61
 +
|align="right"| 70.67
 +
|align="right"| 70.73
 +
|align="right" style="background-color: #cfc; color: green"| '''+0.06'''
 +
|align="right" style="background-color: #ccc; color: black"| '''+0.12'''
 +
|align="right" style="background-color: #cfc; color: green"| '''+0.06'''
 +
|align="right"| 500
 +
|-
 +
| ''DrussGT vs Ascendant''
 +
|align="right"| 57.85
 +
|align="right"| 57.68
 +
|align="right"| 58.11
 +
|align="right" style="background-color: #ccc; color: black"| '''-0.17'''
 +
|align="right" style="background-color: #ccc; color: black"| '''+0.27'''
 +
|align="right" style="background-color: #fcc; color: red"| '''+0.43'''
 +
|align="right"| 500
 +
| 0.43 is NOT "very" red...
 +
|-
 +
| ''PrairieWolf vs DuelistMini''
 +
|align="right"| 42.40
 +
|align="right"| 44.49
 +
|align="right"| 45.07
 +
|align="right" style="background-color: #fcc; color: red  "| '''+2.09'''
 +
|align="right" style="background-color: #fcc; color: red"| '''+2.67'''
 +
|align="right" style="background-color: #fcc; color: red"| '''+0.58'''
 +
|align="right"| 500
 +
| PrairieWolf has a +1/-1 vibrate mode. Both save data. I think there's just a weird / huge variance in one or both of these bots.
 +
|-
 +
| ''Raiko vs Toorkild''
 +
|align="right"| 59.92
 +
|align="right"| 60.11
 +
|align="right"| 60.00
 +
|align="right" style="background-color: #ccc; color: black"| '''+0.19'''
 +
|align="right" style="background-color: #cfc; color: green"| '''+0.08'''
 +
|align="right" style="background-color: #ccc; color: black"| '''-0.11'''
 +
|align="right"| 500
 +
|-
 +
| ''Raiko vs Moebius''
 +
|align="right"| 70.93
 +
|align="right"| 70.53
 +
|align="right"| 72.43
 +
|align="right" style="background-color: #ccc; color: black"| '''-0.40'''
 +
|align="right" style="background-color: #fcc; color: red"| '''+1.50'''
 +
|align="right" style="background-color: #fcc; color: red"| '''+1.90'''
 +
|align="right"| 1,000
 +
|
 +
|-
 +
| ''Toorkild vs Moebius''
 +
|align="right"| 65.54
 +
|align="right"| 65.26
 +
|align="right"| 65.42
 +
|align="right" style="background-color: #ccc; color: black"| '''-0.28'''
 +
|align="right" style="background-color: #ccc; color: black"| '''-0.12'''
 +
|align="right" style="background-color: #ccc; color: black"| '''+0.16'''
 +
|align="right"| 500
 +
|-
 +
| ''DrussGT vs Diamond''
 +
|align="right"| 56.16
 +
|align="right"| 56.19
 +
|align="right"| 56.42
 +
|align="right" style="background-color: #cfc; color: green"| '''+0.03'''
 +
|align="right" style="background-color: #ccc; color: black"| '''+0.26'''
 +
|align="right" style="background-color: #ccc; color: black"| '''+0.23'''
 +
|align="right"| 500
 +
|}
 +
 +
== Bots ==
  
 
Bot versions used:
 
Bot versions used:
Line 188: Line 283:
 
* Marcelo.Alpha.[[DevilFISH]] 1.3
 
* Marcelo.Alpha.[[DevilFISH]] 1.3
 
* [[Barracuda]] 1.0
 
* [[Barracuda]] 1.0
 +
 +
__NOTOC__
 +
 +
[[Category:Research]]

Latest revision as of 22:24, 13 August 2009

1.7.1.4 Alphas

One of the changes slated for Robocode 1.7.1.4 is a new updateMovement() method, which determines the velocity of each bot based on their commands and the rules of Robocode. I'm running some extensive tests on the 1.7.1.4 Alphas and 1.6.1.4 so that we can compare. The main discussion about these changes has been taking place at Talk:Robocode/Game Physics, while some discussion of these tests is going on at User talk:Voidious/Robocode Version Tests.

The score % is for the first bot listed.


Alphas 2, 3, 4

  • Alpha 2 uses the old decel through zero rules (can go from -1 to 1 or -0.1 to 1.9).
  • Alpha 3 uses the proposed new rules (splits the tick into decel and accel portions, eg -1 to 0.5 or -0.5 to 0.75).
  • Alpha 4 caps maximum velocity at 1 when decel through zero (-1 to 1, -0.5 to 1, -0.1 to 1).
  • Turns out all these Alphas contained an inconsistency compared to 1.6.1.4 rules, so they were scrapped for more Alphas and more testing.
1.6.1.4 Alpha2 Alpha3 Alpha4 1.6 & A2 1.6 & A3 1.6 & A4 A2 & A3 A2 & A4 A3 & A4 Battles Comments
Diamond vs Barracuda 99.51 N/A 99.52 N/A N/A +0.01 N/A N/A N/A N/A 500
DrussGT vs Barracuda 99.15 N/A 99.15 N/A N/A ±0.00 N/A N/A N/A N/A 500
Diamond vs DevilFISH 96.73 N/A 96.79 N/A N/A +0.06 N/A N/A N/A N/A 500
DrussGT vs DevilFISH 99.07 N/A 99.07 N/A N/A ±0.00 N/A N/A N/A N/A 400+
Diamond vs Komarious 70.61 70.38 71.08 70.67 -0.23 +0.47 +0.06 +0.70 +0.29 -0.41 500
DrussGT vs Ascendant 57.85 58.04 58.25 58.06 +0.19 +0.40 +0.21 +0.21 +0.02 -0.19 500
PrairieWolf vs DuelistMini 42.40 45.92 47.96 42.06 +3.52 +5.56 -0.34 +2.04 -3.86 -5.90 500 PrairieWolf has a +1/-1 vibrate mode. Both save data.
Raiko vs Toorkild 59.92 60.20 60.18 60.03 +0.28 +0.26 +0.11 -0.02 -0.17 -0.15 500
Raiko vs Moebius 70.93** 73.05 71.71 72.60 +2.12 +0.78 +1.67 -1.34 -0.45 +0.89 500 ** 1000 battles
Toorkild vs Moebius 65.54 65.40 65.54 65.42 -0.14 ±0.00 -0.12 +0.14 +0.02 -0.12 500
DrussGT vs Diamond 56.16 56.26 56.58 56.25 +0.10 +0.42 +0.09 +0.32 -0.01 -0.33 500



Alphas 8 & 9

  • Alpha 8 uses the old decel through zero rules (can go from -1 to 1 or -0.1 to 1.9).
  • Alpha 9 uses the proposed new rules (splits the tick into decel and accel portions, eg -1 to 0.5 or -0.5 to 0.75).
1.6.1.4 Alpha8 Alpha9 1.6 & A8 1.6 & A9 A8 & A9 Battles Comments
Diamond vs Komarious 70.61 70.67 70.73 +0.06 +0.12 +0.06 500
DrussGT vs Ascendant 57.85 57.68 58.11 -0.17 +0.27 +0.43 500 0.43 is NOT "very" red...
PrairieWolf vs DuelistMini 42.40 44.49 45.07 +2.09 +2.67 +0.58 500 PrairieWolf has a +1/-1 vibrate mode. Both save data. I think there's just a weird / huge variance in one or both of these bots.
Raiko vs Toorkild 59.92 60.11 60.00 +0.19 +0.08 -0.11 500
Raiko vs Moebius 70.93 70.53 72.43 -0.40 +1.50 +1.90 1,000
Toorkild vs Moebius 65.54 65.26 65.42 -0.28 -0.12 +0.16 500
DrussGT vs Diamond 56.16 56.19 56.42 +0.03 +0.26 +0.23 500

Bots

Bot versions used: