[Home]David Alves/Phoenix

Robo Home | David Alves | Changes | Preferences | AllPages

Difference (from prior minor revision) (no other diffs)

Changed: 137c137,139
* Hehe, DrussGT 0.3.10 would clean the clock =) -- Skilgannon
* Hehe, DrussGT 0.3.10 would clean the clock =) -- Skilgannon

* Now that would be an interesting change of pace -- Baal

Phoenix


Recent changes
(For full version history see /PhoenixVersionHistory)
  • 1.02 20071102 - ?
    • Stats are rolled slightly faster in AS gun and surfing.
    • Fights a little further away than 1.01 (but not as far as 0.905).
  • 1.01 20071101 - 2122
    • Re-enabled dynamic distancing
    • Back to old WallSmoothing limits.
  • 1.0 20071031 - 2090
    • Major refactoring and speed optimizations... runs about 30-40% faster.
    • Changed some of the segmentations used in the surfed gun and flattener.
    • Changed the WallSmoother?. Currently uses true NoFearTheWalls. - there is no WallSmoothing limit at all. Will change this later, but it's fun to watch. =)
    • Accidentally disabled dynamic distancing in this release.
  • 0.905 20071018 - 2115
    • Removed one gun segmentation and changed another one.
  • 0.904 20071010 - 2114
    • Re-release of 0.903 to see if a robocode bug that was changed in 1.4.7 had any effect on rating.
  • 0.903 20071006 - 2118
    • New flattener, and slightly modified rules for enabling it. I do not expect this to gain me any points.
    • Logs flattener data so that I can do a little analysis and make 0.904 stronger. =)
  • 0.902 20071004 - 2118
    • The new type of segmentation added in 0.901 was not the one I intended. Replaced with the correct one.
  • 0.901 20070930 - 2114
    • Increased the number of buffers surfed from 15 to 24, added another type of segmentation to surfing.
  • 0.9 20070923 - 2109
    • Greatly increased the number of buffers that Phoenix surfs (to 15)
    • All buffers weighted equally (rather than higher weights for highly segmented buffers as before)
  • 0.856 20061121 - 2101
    • Finally a member of The2100Club!
    • More changes to VirtualGuns gun selection code.
  • 0.856 20061121 - 2099
    • Fixed a bug where there was never any bias towards using the main gun.
  • 0.855 20061121 - 2091
    • Changed method used to set bias towards using main gun.
    • New method has larger main gun bias for enemies that are easy to hit.
  • 0.854 20061121 - 2099
    • Increased bias towards using main gun even more.
  • 0.853 20061121 - 2098
    • Better data file compression, now stores data on more bots.
    • Increased bias towards using main gun.
    • Lowered desired distance against enemies that fire high powered bullets.

Scores of Past Versions

MC2K6:
Phoenix 0.56 David Alves WS 99.90 98.54 96.024 98.16 37.82 40.814 66.06 31.78 46.218 60.73 3 seasons except WSC
Phoenix 0.31 David Alves WS 99.95 99.55 98.66 99.39 31.78 37.42 63.31 27.10 42.61 57.93 5 seasons except WSC

TC2K:
Phoenix 0.32 David Alves GF 99.54 74.36 89.08 77.75 77.71 89.36 91.70 87.71 87.75 92.90 86.79 35 seasons, 0.182
Phoenix 0.61 David Alves GF 99.44 73.57 88.49 79.78 76.57 91.61 91.82 87.54 88.42 92.34 86.96 166 seasons, 0.088 (no joke!)


Chat: (Archived chat is here)

Seems you will keep Voidious company on the other side of the once untouchable border. Now just wait till it is stable . . . -- GrubbmGait

Not with 0.852. Maybe 0.853 will do it. You've definitely got the name right though. =) --David Alves

Exciting! *Fires up clients again.* -- Voidious

Wow, this could be a nice new turn. -- Chase-san

So far (700+ battles, 2105 rating) it has only lost to Shadow. It beat Dookious with 53%. -- Martin

Oh, the terrible pain! 2098.18, 1167 battles. There's still time, though - Dookious 1.26 was at 2104 with 1000 battles. Still got my fingers crossed for ya! -- Voidious

So I was pitting Phoenix against RabidWombat (just for kicks and giggles, mind you, not with any serious idea that my bot would actually win), and just out of curiosity I tried turning on Phoenix's debug graphics. What a wondrous sight met my eyes! Your debug graphics are not just functional, they're purdy. I was curious as to what they all mean, though. Some of it was obvious, but I'm rather mystified by the numbered color bars on the right. Care to enlighten us? -- RobertWalker

Sure. The bars on the right represent each of the buffers that Phoenix is surfing. They range from least segmented at the top to most segmented, the 3rd one from the bottom. The colors range from blue (least dangerous part of buffer) to red (most dangerous part of buffer). The number next to the bar says how dangerous pure red is for that bar. The second one from the bottom is the sum of all those buffers, and the bottom most one is the sum after significant smoothing. The incoming waves are color coded using the colors from this bottommost buffer. The pink dots that show the two paths that Phoenix is choosing between (clockwise vs. counter-clockwise) when it is surfing. Sorry for the slow response. =) --David Alves

Whatever you gave to Phoenix 0.9, I want some =). 2108 at 1500+ battles. -- Skilgannon

Lighter fluid. ;) --David Alves

Looking at it's detail sheet, Dookious seems to like this release =). Not quite as strong in the PL, but stronger in the main rankings. So I'm guessing that it is more reluctant to use the surfer gun and to enable the flattener. Nice work! I personally think propane is better than butane, though. It's lighter than lighter fluid =). -- Skilgannon

I didn't change the code for enabling the flattener or the code for selecting which gun to fire with. The only changes were increasing the number of buffers that it surfs and making them all weighted equally. --David Alves

Cool! I'll have to try weighting the lower buffers. Currently, once I have a hit in a higher buffer, I ignore the low buffers entirely. -- Skilgannon

Held my breath when I checked the rankings this morning. =) Great work, dude. Might have to put the Lukious/Rewrite on hold for a bit... -- Voidious

Weel done, only have to sqeeze a few more points out of it. ;) --Chase-san

Looks like your new flattener does you well in the PremierLeague! -- Simonton

Well so much for that... I called it 1.0 after I saw what how badly it beats Dookious, but the changes seem to have cost me points overall. Oh well. Here were the test results that prompted this release:

 1st   davidalves.Phoenix 1.0       2893   1050   210   1419   214   0   0   21   14   0
 2nd   voidious.Dookious 1.554NDS   2151    700   140   1182   128   0   0   14   21   0

 1st   davidalves.Phoenix 1.0       2940   1100   220   1417   203   0   0   22   13   0
 2nd   voidious.Dookious 1.554NDS   1933    650   130   1030   123   0   0   13   22   0

 1st   davidalves.Phoenix 1.0       2851   1000   200   1446   205   0   0   20   15   0
 2nd   voidious.Dookious 1.554NDS   2283    750   150   1239   144   0   0   15   20   0

 1st   davidalves.Phoenix 1.0       78179   29400   5880   37506   5393   0   0   591   409   0
 2nd   voidious.Dookious 1.554NDS   56633   20100   4020   29258   3254   0   0   409   591   0

NDS stands for no data saving; Phoenix's data saving was also disabled.

 --David Alves

I want to see some new MC2K7 results.... you're knocking on the door! -- Skilgannon

No thats the sound of Dookious slamming the door shut on Phoenix's foot. :P --Chase-san

So, as of 1.31, N is the only bot besides Shadow to beat Phoenix. What do I win? --Baal

It's because of the funny versions of the rumble at the moment, I'm quite sure. Try running them both in 1.4.9. -- Skilgannon

Eh, guess I did speak too soon. But, I can still smile at the screenshot :) --Baal

Haha, yeah. It shows that if we were limited with the amount of CPU time allowed, things would be very different in the rumble =) -- Skilgannon

This makes me wonder if it might be nice to have a CPU-limited rumble class as well as the size classes ;) -- Rednaxela


Robo Home | David Alves | Changes | Preferences | AllPages
Edit text of this page | View other revisions
Last edited February 13, 2008 6:54 EST by Baal (diff)
Search: