suggestions

I respect Enders thoughts on this but I think it's about time the tournament changed to best out of 3 or 5 fights. Winning from luck has been fun for a while but I would like to see a result best out of 3 or 5. I don't mind if it's only for one tournament only. I think this would be interesting and might even convince more of the player base to take part. Thank you ^^

I hate to suggest things since I am not a proper active player but I love tournaments :P ' I know this had been brought up many times '

My 2015-04-02 post here still accurately describes my thoughts on this:

http://bots4.net/forum/4/6509

I like the way it is. Yes, it depends on the luck but not that much because you need a bot that wins atleast 40% of the times against the best bot in that cat. I would say that a bot should win more than 45% of the times if it really wants to have a chance at winning.

For me, the 3 main factors determining the chances of winning are (by order):

1- Build

2- The bots that sign up (ie, if there is lots of bots with similar builds or kinda good builds that could give a loss to the best build)

3- Luck (randomness) becomes the main factor when there are lots of bots with the same build.

Regarding category 2 (251-300), it's without any doubt the most competitive cat. Every bot that signs up are kinda good. Ofc ReneDescartes, Ardor and Dyna are the top builds (they come top3 everytime) because the other bots are optimized to be either ratio whores or energy dumpers. Now that Dyna leveled to 300 in theory it would have more chances than Ardor or Renedescartes that are at 290/292. However it didn't won this edition. I'm pretty sure if the tourney was 3 fights instead of 1 Dyna's chances of winning would decrease. Simply because that build isn't the best at lvl 300 and that's the most important factor.

At this point you may as well just suggest "Hey Ender, can you update the game kinda...at all in any meaningful way?" and you'd get the same replies as this thread.

The tournament are screwed and that's all there is to it. Due to lack of human interaction, the results will always be fixed to a degree. It really doesn't matter if 2 blasters with the same build go 1 time or 3 times, the difference in fights will vary only a slight bit more. Only if it were 1,000 or more fights would it make a large difference and produce a more "this bot won for sure, 100%" result.

I disagree... My lvl 61 bot has a very low chance of winning anything, and often only win 1/18 battles, if even any. Never the less I managed to gain 1 point one time, by winning 3-4 battles or so. That was probably pure luck... The chance of that would be smaller if it was just 3 battles.

I think it would be fun if the amount of battles varied from 1-10 or so, making some tournaments based on builds and some slightly more based on luck

Tournaments work perfectly fine. Rene and Myriad built bots with the best or very close to the best build for each cat. What happens? One of their bots usually win. ALWAYS. When neither of their bots win is because there is another bot(s) with a similar build.

If the bots have similar builds shouldnt be the winner based on luck?

Now, 100, 1000 fights for what? To make sure my simulator is working fine and Rene and Myriad win all the time? Pfff

1 fight is the only thing making worth to watch the results of each tournament edition and even so they are almost always the same.

With all respect to ESV, I don't think they would win all the time at all. Once they're all 300, it would probably be best to give them a full 400 fights to find out the outcome. I just have a lot of faith and belief in my bots, so I like to see things done other ways so I can maybe improve it by making another bot ^^ Rene being in my clan means I can't test my build to see what needs improvement to win a tournament. No way will I leave Escapism to find out. Rene and Myriad might win more often than me for sure but 3 or 5 fights would certainly be good in my opinion. Anyway. Just a suggestion so I'm happy with how things are just thought I'd mention my thought. Also never found out how to work the bots simulator so this is the only way lol ^^

The more fights you have, the less impact RNG can have on the outcome. 1 fight to decide the winner gives us the possibility of upsets, whether that's a good or bad thing is a different discussion entirely.

Damn this luck -_-

Better luck next month ^_^

I like the tourney the way it is. I just wish I would remember to enter... :)

would still like to see more fights, a best of 5 system would be nice, cant complain though, im still rocking it

I am attacking each bot that stays 300; and doing best of 5 just once and I have yet to lose, but I know I will at any time. Unpredictable results daily it seems. Tourney is as people say a game of luck. :P Congrats Otto! I thought Lyrad might have won. Totally agree about the best out of 5. That would be awesome. I hope to be the lucky one next month ^^

Much respect

cat 18, 17, 16.. 2nd, 3rd, 4th places all on the same wins, that doesn't show who has stronger bot. we definitely need more fights!

based on average con left over

tbh the tourney is utter bullshit the way it is iv attacked everybody with dynamitecop that i am able too and beat them easily first to 5 and yet that bot as failed to reach first its the best bot over all and still wins jack shit

i use to look forward to it and i use to use the tourney as my way of seeing if my bot as improved over the levels/month obviously when it was bots2

it really needs putting to out 3-5 fights at least

I'm curious whether people's claims on this thread that the variance is too high can be backed up with data. I have the ability to easily simulate battles between 2 bots (using the exact same code as used by both tournament and non-tournament battles), so if someone can point out a bot they think is consistently getting "unlucky" where it's seemingly the best bot in the category, but keeps losing to due to the purportedly high luck factor, then as a challenge, I'd like to see if I can disprove this anecdotal evidence with hard data/simulation.

I think you had one just above

Btw I just want to state; although you probably already know. I let Satanic use my bot before, so he has had experience fighting with it.

Ender, I think you could simulate maybe just the top 5 bots to see how it works out? I've only ever had confidence in my Bots while others have different opinions on it. I always feel like doing an entire 400 attacks just to see how it would work out, but I respect people too much to do so. If I knew how to simulate myself I wouldn't have even bothered making this thread. I hope you can find out how it would result and post with your thoughts? This would I'm sure put an end to this " That bot was damn lucky winning; that would never happen in 3 or 5 fights. The one fight per tourney might be accurate least this way we will know for sure. Too much difference of opinion on this.

Much respect

I myself would test it on Usernames Dynamitecop, Ardor, Shattered, ReneDescartes

as all of them have won the tournament other than Dynamitecop

Thank you for finally testing this out Ender,

Also would probably best i you fight them all against Dynamitecop

http://i.imgur.com/k3W0iJH.jpg

Hmmm if in tournaments would be more fights between each bots, maybe it will be more accurate to see the winner.. I would like to see how my bot goes 1on1 vs 1st place winner in total of 50/100 fights. I could do that by myself, but then i won't be able to compete in next tournament because of gained exp..

the question is, Will ender be testing this out?

Will you be testing this Ender? or is it one of them things that is just put on the list to do?

This hasn't fallen off my radar, but I'm on vacation now and so it'll be a week or two until I have a chance to play around with this. Sorry for the delay.

Oh, enjoy your vacation ^^ And no rush, it was just an idea to see how things would turn out. I am glad you're going to look into it. I look forward to finding out the results later on.

http://imgur.com/a/P8xwU

just a picture of dynamitecop beating shattered first to 10

This game just doesn't want me to taste success lol

Now if ender doesnt think the tournament needs changing then something is wrong

TheCause came 5th FIFTH wtf? i even finished above him something needs doing soon

Ender, are you actually testing this or is it one of them things you have put on your to do list and it will get done when ever you can be arsed?

Btw Ender, you can also test Cat 3. By my simulator, I should win against Rubor 58% of the times, which in tournaments usually result in frequent wins. You can see I´m the bot with more HP remaining from Cat 3... lets see if this year I have a christhmas gift from bots!

This is still on my radar. I'd been working on the treasury bug first because there was clear evidence of a problem in that case, whereas in this case it hasn't yet definitively been proved there is a problem. It's still top of mind for me to dig into as soon as I have a chance. Sorry for the delay and thanks (as always) for your patience.

Esv, will do.

No problem, Ender. Hope you get the chance to solve any problems you might find. ^^ Cheers

so let me get this straight ender, you worked on a treasury bug that only escapism used... which they have lyrad thecause myriad training every day filling the treasury... so pointless really but instead of working on something a lot of people are bothered about and everybody can see there is a problem you decide to work on a shitty little bug

thecause finished below me in the last tourney in what way is that not proof that the touney is fucking shit?

dynamitecop never wins a tourney yet he beats them all in a race of 10 and wins by a quite big margin

No wonder the game is going down the shitter ender... the problem isnt just the game the problem is you and the lack of updates and the updates we do get a fucking shit such as a treasury bug and changing words woop woop

Have to admit I felt a bit cheated at the last tournament as one of the two bots I have bothered to build for it came 3rd in the lvl91-100 cat, yet I had a LOT more Hp remaining than any other bot! :(

But hey ho it's only a game, sj swearing is slightly cool in moderation and your ungreatfulness isn't constructive... Seen as its Christmas (the time of giving) maybe you can give Ender some love and tell him all the things you like about him :)

tl;dr: I picked out three cases from this thread for investigation. Although there are definitely cases of underdog bots coming out on top, this possibility for variance is by design. Unlike with the equipment condition tournament bug from last year, in this case I was unable to find any instances of statistically improbable events, so I feel comfortable saying everything appears to be working as intended.

Note: I'm not against some sort of "every bot fights every other bot 10 times instead of 1 time"-style tournaments for the future as a "best of the best" infrequent special edition or something to appease those that don't like the RNG. The tournament code would need to be reworked quite a bit, but I'm (more generally) in favor of there being more variety to the game. For now though, the tournament will remain as the "one battle per pair of bots"-style, which comes with increased variance.

Investigation #1 - Esvrainzas vs. Rubor (edition 22, category 3)

Claim: Esv's calc says he should beat Rubor 58% of the time, which usually results in a tournament win for him.

Data:

$ php src/php/util/simulator.php 22 13681 22 12924 10000
Attacker: Esvrainzas [250] (id 13681) - tournament 22
Defender: Rubor [250] (id 12924) - tournament 22
Progress: 10000/10000 (100.00%)
Elapsed: 127.406 seconds
Attacker wins: 6023
Defender wins: 3977

Analysis: My simulation pretty closely matches Esv's expectations. However, a 40% chance of losing is significant. In this case, the 40% chance event happened over the 60% chance event. This win for Rubor over Esv allowed it to edge it out in the rankings, making Rubor the category 3 winner.

Commentary: This is a good example of why I prefer 1 battle per pair of bots. To me, it's less interesting if the bot that wins 60% of the time is essentially guaranteed a win. If it's a 60/40 matchup, I think it's okay for the underdog bot to sometimes win. If there were, say, 5 battles per pair of bots, the chance of the underdog bot winning all 5 battles would be a pitiful 1%.

Investigation #2 - Dynamitecop vs. Ciuc (edition 22, category 2)

Claim: (not so much a claim with this one, but more just general curiosity about matchups/variance)

Data:

$ php src/php/util/simulator.php 22 11 22 433 10000
Attacker: Dynamitecop [300] (id 11) - tournament 22
Defender: Ciuc [289] (id 433) - tournament 22
Progress: 10000/10000 (100.00%)
Elapsed: 98.044 seconds
Attacker wins: 7889
Defender wins: 2111

Analysis: This is the 2nd place bot (Dynamitecop) vs. the 12th place bot (Ciuc) for category 2. The 12th place bot beat the 2nd place bot in their individual matchup.

Commentary: As with above, underdog results can and will happen. The larger win percentage chance in this case is consistent with how large of a difference in overall tournament rank there is between these bots. Although the weaker bot won the individual matchup because of variance/luck, the fact that there are many battles for each bot works against the variance, thus making it much more statistically likely that the stronger bot ultimately finishes higher in the tournament rankings, despite the unlucky loss.

Investigation #3 - Satanus Inaximasus vs. TheCause (edition 22, category 1)

Claim: It should not be possible for Satanus Inaximasus (4th place) to finish ahead of TheCause (5th place).

Data:

$ php src/php/util/simulator.php 22 40 22 142 10000
Attacker: Satanus Inaximasus [332] (id 40) - tournament 22
Defender: TheCause [394] (id 142) - tournament 22
Progress: 10000/10000 (100.00%)
Elapsed: 138.088 seconds
Attacker wins: 1442
Defender wins: 8558

Analysis: The data certainly lends support to the claim. But first off - is this a one-time thing or has it been happening frequently? The answer is that it is a one-time thing - I checked editions 15-21 and in all cases, TheCause finished ahead of Satanus Inaximasus. So that reduces this to showing that a one-off instance of the reverse happening isn't an impossible result. To dig into this further, I simulated both bots against Lyrad (data not shown for conciseness), the second-place finisher. In the tournament, Satanus Inaximasus beat this bot, but TheCause did not. In the simulation, it turns out that Satanus Inaximasus had a ~15% chance of winning, whereas TheCause had a ~54% chance. More unluckiness on both parts.

Conclusion: This case is more extreme than the other two cases, but not impossible. In any tournament, there are always going to be examples of underdog bots winning, sometimes in more surprising cases than others. If it were a regular occurrence that a bot with a 14% chance to beat another bot was always finishing ahead of that bot in the rankings, that would be reason for investigation, but in this case, it looks like a one-off unlucky event.

RIP SJ, you just got yourself enderfied!

Thanks for looking into this ender... I think Both BO1 and BO10 has its pros and cons... Having both best of 1, 3, 5 and/or 10 every once in a while is no harm =)

BIG thank you to Ender! Bring on tournament Christmas! ^^

Dyna you still wont win even though you have the strongest bot mate as its wank

Patience, patience. Hopefully one day