[Home]Paul Evans

Robo Home | Changes | Preferences | AllPages

Creator of the all round bot SandboxDT, also SandboxLump and SandboxMini (plus other Sandbox series bots).

Click [here] for details of the VirtualBullets class used in SandboxLump and information on rating robots.

Come to /Inspiration to thank Paul Evans for his contributions and innovations in Robocode!
I am curious now. I thought my new movement was ready for prime time. Gouldingi 1.5 certainly gives a few top bots a run for their money. But. There's a huge but. DT just eats the new Gouldingi and spits it out. There's obviously a weakness to the movement. But I can't see it with FloodGrapher and Vic's ZoomTargeting seems unable to exploit it as well. Do you see what it is with DTs stats? If you do, could you please tell me? -- PEZ

Actually, while Gouldingi does reasonably well against Fhqwhgads, it still seems to lose pitifully to FloodMini. They may not have the real answer you want, but it may be worth looking closer to see what more segments look like. Gouldingi's wall movement, btw, now officially kicks the crap out of mine (although it looks similar to something I tried the other day). -- Kawigi

Hmm, FloodMini runs in 40 fps on my machine so I can't really test extensively. But in the 100 rounds battle I did test Gouldingi won 65-35 in survival and even bigger in score. But let's assume your tests show the correct picture, 100 rounds is way to little to trust. What more segments are FloodMini using that Fhqwhgads aint? In larger tests Gouldingi seems to loose only by a small margin against FloodHT as well. Is FloodMini using any special segmentation that HT doesn't? I would really like to solve this mystery. My only lead proved to be a false track. Good to hear the wall movement is worth anything. It's nothing like DT's but it's probably much simpler and simplicity is worth a lot when improvement work is ahead. I'll reveal the method when the ER is up and running again and we have some new rankings to know how competetive it is. -- PEZ

No, the segmentation between HT, Mini, and Fhqwhgads is all almost identical. FloodMini (and maybe Fhqwhgads) define the wall segment differently than HT (the new FloodMini will do it the way HT does), and HT segments on bullet travel time instead of distance (but Fhqwhgads 1.1 and the next FloodMini will do that as well). However, they have one key weakness if you're testing against them a lot. Do you always delete their data before testing a new tweak of the movement? They are all really bad at picking up changing movement, since none of them roll their averages at all (the next HT will fix this possible vulnerability). The other weakness they may have is that they assume you don't react directly to bullet fire (beyond possibly remembering the power). Gouldingi doesn't appear to do this anyways, though. And Fhqwhgads seems to be more vulnerable to Gouldingi's targeting system than the current FloodMini is. Here is FloodMini 1.3 with no pre-saved data on Gouldingi 1.5:

1st: kawigi.sbf.FloodMini 1.3	8767	3250	650	4234	628	4	0	66	35	0
2nd: pez.mini.Gouldingi 1.5	5553	1750	350	3086	322	2	41	35	65	0
And, for good measure, with that saved data advantage:
1st: kawigi.sbf.FloodMini 1.3	11022	4300	860	4945	916	0	0	86	14	0
2nd: pez.mini.Gouldingi 1.5	3378	700	140	2404	133	0	0	15	86	0
So my guess is either that Gouldingi is just completely firing in the wrong place for FloodMini, or that you should be able to find something in there that's weak. Do you run FloodGrapher with the -v flag on? That might help you narrow it down (as that's what I use to 'debug' movement) -- Kawigi

Anyone think Paul Evans will come back? Come to think of it, is everyone quitting robocode or what? -- Mike

I'm sure he'll be back. RL catches up with all of us sooner or later. -- Tango

I'll be back, but not for some time - I have to re-build my garden - I have 100 railway sleepers about to be delivered. Out of interest has anyone generated a matrix of performance for say the top 10 or 20 bots over say 1000 or 5000 rounds - I think DT may be rather good at that competition. -- Paul

It definately would. I have measured it in that neighbourhood quite a few times lately and it comes second only to RaikoMX then. I might have some time this weekend to make a more formalized test. Even though I too have a garden to attend to. 100 sleepers! You must have a large garden. -- PEZ

The back garden is not massive (80'x40') but there is a height difference of 3 feet from front to back. I am going to have 6 runs of sleepers so that my grass/patio areas are in several flat areas (at the moment there is just enough slope to make using garden furniture a real pain). I've just bought a cheap digital camera - you may see a picture of me, the garden and the sleepers in the WhatWeLookLike page in a few weeks time. -- Paul

Hmmm, I can't really deal with Brittish measures in my head. My garden is 1004 m2. Of Which the house takes up about 100 m2. How does that compare? I've built a quite massive deck (75 m2) to deal with our need for flat ground. The rest of the garden is left quite wild (with pine trees, oaks, rocks and stuff). The height differences is something like 5 meters. I would need to buy a bit more than 100 sleepers to even it out. =) But I'm in the process of ordering some sleepers to build a terassed lot where we can grow some potatoes, carrots and what have you. "We" is my wife I guess. I do the digging and carrying and suh and she tends to the stuff that grows. She's got green fingers. (If that's an expression that works in English...). The good news is that I spent quite a lot of time in working with this garden last summer, but still had time to participate in the Robocode race. -- - PEZ

If my maths is sound, Paul's garden is just under 300m^2. -- Tango

What are sleepers? Some sort of plant? -- Alcatraz

Railway sleepers. It's what keeps the two rails together. On some modern railways the sleepers are made of concrete. But in good'ol railways they are made of thick log-wood. Heavily prepared with some oily substance making them resistant to weather. They make excellent little boundaring walls in gardens so that you can fill with soil and level height distances out. (Sorry if this doesn't make too much sense. Both gardening and construction are outside the domains where I have an English vocabulary ...) -- PEZ

Your English seems good enough to me. We use sleepers as steps from the patio to our garden. -- Tango

Thanks, it's just that things like "Heavily prepared with some oily substance making them resistant to weather" probably have words of their own. In Swedish it would be "oljeimpregnerad". -- PEZ

That looks like the english word "impregnated", which would be correct. "Treated" is the other word used to describe it. -- Tango

BTW talking about swedish expressions. Do you know something to eat which is called surstromming (or something like that). It should be a swedish delicacy of old fish??? --deathcon

Indeed. Surströmming is one of my favs. Though I would guess only about 10% or so of us Swedes like it, 10% can eat it and 80% totally and utterly hate it. I can't explain the cooking process in English, but the direct translation of surströmming would be "sour herring" and that's a clue. It's very salty too. I guess this kind of food exists in many places. In Iceland I was introduced to "Surhaj" (sour shark) which is produced about the same way. And I saw on TV some years ago that the Innuits cook something like "sour seal". It makes the food able to store for very long periods of time. Good in older times when the food supply could be very arbritrary. -- PEZ

I just ran across an [interesting article], and I happened to notice the author's name. Coincidence? -- Skilgannon

Robo Home | Changes | Preferences | AllPages
Edit text of this page | View other revisions
Last edited April 1, 2009 17:50 EST by Skilgannon (diff)