[Drownhim.com] Testing Unibot settings and sharing successful bots

[LoyceV]

A month ago, I discovered Drownhim.com:

Since Rollin.io (where I was a Mod for more than 2 years) closed, I haven't really used any other gambling sites. Drownhim is more of a multiplayer game than a gambling site, and that's what I like about it.
This thread is specifically made for testing Bot settings. Feel free to post your own settings, and/or try to beat mine. For testing, I use the room called "Bot training":

Unibot
Drownhim offers manual playing, and several Bots. The most interesting one, and the reason I created this topic, is Unibot:

I'll summarize the Unibot settings, the above example would be written as: (1,3+ 75%; 1 30%; 0 20%; 10 7 3 9).

Bot testing
For testing bot strategies, Admin helped me to create 9 different accounts. "Normally", using more than one account is not allowed, just like in online poker. For testing Bot strategies, Admin made an exception for me.
These are the accounts:

Code:

DeepBlue
T-1000
R2-D2
Wall-E
Android
HAL9000
System_370
Apple_II
PC_AT

For testing, I'll start with some settings, then run 6-9 Unibots for about 50-100 rounds, and see which one wins most:
Green: best result(s)
Red: worst result(s)
I'll change the settings from the worst performing bot, and try again.

No spam
All my threads are now self-moderated to stop signature spam. I will remove all irrelevant posts. If you quote the entire OP, your entry will be deleted.

Disclaimer
I'm not affiliated with the site, but I've been in contact with MrDrowner, and may run one or more campaigns for them in the future.

[1713515785]

1713515785

[1713515785]

1713515785

[1713515785]

1713515785

[1713515785]

1713515785

[1713515785]

1713515785

[1713515785]

1713515785

[LoyceV]

Reserved

[LoyceV]

Test bot settings #1 (6 bots):

T-1000 is brutal: drown everyone it can, 100% of the time (0+ 100%; 0+ 100%; 0+ 100%; 10 7 3 9).
Android is lazy: running default settings (1,3+ 75%; 1 30%; 0 20%; 10 7 3 9).
R2-D2 only retaliates after 2 drowns in 3 moves (0+ 50%; 0+ 50%; 0% 5%; 10 7 4 11).
DeepBlue is supposed to be the smartest, but that failed (1+ 100%; 0+ 50%; 1+ 40%; 10 7 3 6).
Wall-E is sweet (0+ 40%; 0+ 30%; 0+ 80%; 10 7 3 30).
HAL9000 only drowns 50% of the users, but strikes back hard: drown him once, and he'll drown you 3 times (0+ 50%; 0% 50%; 0% 50%; 10 7 3 1).

[MrDrowner]

We’ve conducted another test of bot’s strategy with the goal to determine which value of “Revenge factor” leads to winning. For that, 6 bots were run in the “Training bot” room for 145 rounds.

All of them were programmed to drown another on Upper Level on 1+ moves with probability 30%, and on My Level on the first Move only with probability 20%.

The penalty points were:
 for drowning me at Move-1: 10
 for drowning me at Move-2:  7
 for drowning me at Move-3:  2



The bound, at which the bot “revenge” specific player for drowning himself were different for all 6 bots. The results of this run are as following:

UserID	Revenge bound	Gain/loss
Mosquito	1 (revenge if he drowns me at any of 3 previous moves (most rancorous strategy)	-1.38
Beatle	6 (revenge if he drowns me at any of 2 previous moves)	1.99
Dragonfly	8 (revenge for drowning me either on last move or (move-2 and move-3))	0.71
Fly	11 (revenge for drowning on last move and on (move-2 or move-3))	4.88
Ant	16 (revenge for drowning on last move and move-2)	-2.37
Grasshopper	18 (revenge for drowning on all 3 previous moves (most forgiving strategy)	-5.64

   

The above bots keep running on the “testing bots” room. Anyone welcome to beat them - e.g. by improving the most  effective strategy (Fly).

The logs for the above 145 rounds are available by clicking "full log" and choosing "Bots training" and staring time 9:42 UTC July 11, 2018.

[Gaaara]

Wow, I really enjoyed how they made it simple but so fun I already have a strategy and the good thing about this game was everyone has a chance to win some. Having a name that's easy to see would be the worst thing to do though if having a name with less characters and less appearance will make your chances higher. I really liked it, it was fun simple and entertaining.

[MrDrowner]

Yesterday the new bots features has been released: now everyone can save a number of his strategies and share it with others if he like.

"Insects" bots has public strategies  which anyone can see in the strategies list. They differs from each other only by first param - probability to drown players at upper level:




The results after 213 rounds in "Bots testing" room are:

UserID	Strategy name (meaning of the numbers see above)	Gain/loss
Grasshopper	50-70-11	-5.00
Mosquito	60-70-11	-0.37
Beatle	70-70-11	-5.05
Dragonfly	80-70-11	3.18
Fly	90-70-11	-0.44
Ant	100-70-11	3.07

As one can see, Fly result of -0.44 is quite unexpected between +3.18 and +3.07. It means that it's not so simple as it may looks like..

[LoyceV]

Loyce am i gonna merited for sharing my bots info?

Asking for Merit makes it highly unlikely you'll receive any.

Since the new features were released, this thread is more or less useless It's much easier to share bots on the site, which also shows the success of each bot (as a percentage of the amount won or lost).
The success of each bot largely depends on your opponents. It's not possible to create one bot that always wins. For instance, if all opponents use an aggressive strategy, you can't win if you're not aggressive too. But when 3 aggressive bots play against 3 "peaceful" bots (which means they only Drown players after they were drowned), the peaceful bots win!
That makes it necessary to use different bots for different scenarios, and adjust your bot settings when your opponents change theirs.
At the moment, the best strategy is to win the contest: create a bot that has the best win percentage in 1500 rounds.

[khatarnak]

hello guys.
my username in game is Persian.
finally i am going to share my bots config + pics.
lets start with my account overview.



List of my bots - insects




List of my bots - 25 Moves



when i started this game the most    using bot was peaceful ant - as the name of the bot says it was peaceful type.



and i made simple #336 against them. rate 132% in 610 rounds



the 25 Moves simple config was the same as insects and that was successful too. rate 73% in 103 rounds

in past 2 days the peaceful bot shown named rancorous super coop#470 .it was the best peaceful at the time and with numbers they seems unbeatable.



and today i found solution to beat them all . doesnt matter how many roncorous is in the game.
i build keep it up#561A  rate 188% in 97 rounds.
its propably the smartest bot in the game until now.
in the insects room every 3 move the last line of the game falls. and as peaceful bots doesnt do anything offensive(they just revenge) this bot drown them all on the third move.


I hope this will help you.

[LoyceV]

At the moment, the best strategy is to win the contest: create a bot that has the best win percentage in 1500 rounds.

It worked, I won the contest. See here to read how I did it.

As I said before, the success of any bot depends on the other bots in the room. At the moment, I'm having good results with these bots:


The one on top has these settings:


What it does, is simply wait for the round in which all other bots are too peaceful and don't drown anyone. Then, at the last Move, it drowns all opponents and instantly kicks them out of the game.
These are the results per game:


Why am I sharing this? I like the game! And as much as I like winning, it's more fun with real players. All rodents and insects (example: Bee, Wasp, ...) are Admin bots to keep the game alive. That's why they sometimes use the same strategy.
Long term, the game can only be sustainable with real users. If 3 or more real users join, the Admin bots won't join that round.

[MrDrowner]

What it does, is simply wait for the round in which all other bots are too peaceful and don't drown anyone. Then, at the last Move, it drowns all opponents and instantly kicks them out of the game.

By this Loyce has proved that 1 aggressive bot beats 6 pacifists (there were 7 places in the round that time).

But I'd like to prove that a group of pacifists will always beat an aggressor (here is non-triviality of this game).

Now there are 9 places in the Trial room and all our bots (insects) are pacifists (drown with only 1% probability, but revenge for drowning themself  of course). If someone come with aggressive bot (like Loyce's one) it will climb with 8 pacifists - it's a question if it can beat 8 with the same ease as 6.
Also pacifists will now be added with a number of moderate aggressive bots - with 10% probabilities (mosquito, beatle, grasshopper, ant, fly, dragonfly).

[LoyceV]

Now there are 9 places in the Trial room and all our bots (insects) are pacifists (drown with only 1% probability, but revenge for drowning themself  of course). If someone come with aggressive bot (like Loyce's one) it will climb with 8 pacifists - it's a question if it can beat 8 with the same ease as 6.

I can still beat 8 of them
I'm now up against 8 of these:

I'm using the same settings as before.

With one player like me
The probability of them drowning anyone is 1%. I win if nobody drowns anyone. If just one bot drows another bot,  I'm very likely to lose.
I hope my math is correct:
Round 1: the probability of nobody drowning anyone is (0.99^8)^8 (=0.525596) (because there are 8 bots drowning 8 others with 1% probability)
Round 2: the probability of nobody drowning anyone is (0.99^8)^8
Round 3: the probability of nobody other than me drowning anyone is (0.99^8)^8

The probability of nobody other than me drowning anyone for 3 rounds is ((0.99^8)^8)^3 (=0.145). That means there's a 14.5% chance I win the round if I drown them all on round 3. With a 1 bit buy-in and 8.91 bit pot, that gives me an average return of 1.29x my wagered amount.
There's a small bonus: sometimes I win a shared prize (say 1.11 bit), which slightly increases the returns.

With 2 players like me
If anyone were to join using the same strategy, the math changes:
Round 1: the probability of nobody drowning anyone is (0.99^7)^7 (=0.611117)

The probability of nobody other than us  drowning anyone for 3 rounds is ((0.99^7)^7)^3 (=0.228). That means there's a 22.8% chance we both win if we drown them all on round 3 and have to share the pot. That gives an average return of 1.01677x the wagered amount.

[MrDrowner]

Now there are 9 places in the Trial room and all our bots (insects) are pacifists (drown with only 1% probability, but revenge for drowning themself  of course). If someone come with aggressive bot (like Loyce's one) it will climb with 8 pacifists - it's a question if it can beat 8 with the same ease as 6.

I can still beat 8 of them
I'm now up against 8 of these:

I'm using the same settings as before.

It seems 1% aggressivity in the GROUP is not enough to defend from 1-2 aggressors. Now 6 insects (Cricket, Flea, Ladybug, Moth, Spider and Termite) has increased the aggressivity from 1 to 5 persent, will see how Loyce will beat them..

[khatarnak]

(#1108 AAA) is on fire    about 10 mbtc profit.

[TheBeardedBaby]

Interesting. Trying it now, looks like fun game
I entered the wrong room, i'm in trial (bot fight) but it goes well

[LoyceV]

Interesting. Trying it now, looks like fun game

If only more people joined
It could be so much fun creating bots to fight 10 people all the time, adjusting my bot when other bots change too, making it a continuous battle for improvement.

[KingZee]

Interesting. Trying it now, looks like fun game

If only more people joined
It could be so much fun creating bots to fight 10 people all the time, adjusting my bot when other bots change too, making it a continuous battle for improvement.

I've actually sent a longass e-mail to the drownhim admin, I'm rewriting part of it here to maybe check interest for the idea :

So, first simple idea to add to your bots if you like it (you can also ask other users if they like it):

Divided percentages : Sometimes, I want to drown users under me at 50% chance. But sometimes, in same game, but in different rounds, I want to drown them at 25%, or maybe 90%. I would like to be able to pick multiple chance strategies for different rounds.
For me it would look something like this,

Upper : 1 : 25%
Upper : 2+ : 75%
My Own : 1 : 50%
My Own : 2+ : 25%
Lower : 3+ : 50%

If two of them overlap, you can give latest one priority, or something.

Also this second problem isn't related to the bots, but to profit.
Imagine I start my bot, and write max loss : 2 bits. And imagine my bot made 32 bits profit for 3 hours, and then some other player come and upgrade his bot, and now my bot lose 100% of the time, will I come back to find my balance -2 bits? That's not fair. Best would be to come back and find 30 bits, which means instead of counting total losses, you should count cumulative losses. If I lose 2 bits in a row that's already problem, no need to wait until I lose 32 bits in a row to stop my bot. I hope that makes sense, that feature I'm sure is 100% needed, that's what it should be, if bot loses X bits in a row, stop it, not if balance becomes -X bits, then stop.

So besides the second problem which I'm sure is 100% going to have a positive feedback from pretty much anyone with a little logic, what about the first idea? Would you like to see it implemented? Let the admin know your feedback! (Not through me obviously I'm not affiliated with him, but through the website!)

[LoyceV]

So besides the second problem which I'm sure is 100% going to have a positive feedback from pretty much anyone with a little logic, what about the first idea? Would you like to see it implemented? Let the admin know your feedback!

I'd love to see more bot options, but I'm afraid it'll quickly become too complicated for new users. Therefore it's probably better to wait until there are much more active users before adding new bot-features.
It would be nice if bot settings can depend on the number of players left: if for instance you're 1 out of only 2 players left, the best strategy is to drown the other player all the time. However, he'll do the same, so you won't win much by adding this customization.

I would suggest to add a "take profit" and "stop loss" to the current "max loss". That way, you could for instance set these options:
-max loss: 2 bits
-take profit: 32 bits
-stop loss: 4 bits
So if you instantly lose 2 bits, the bot stops. If you reach 32 bits profit, the bot stops. And if you reach for instance 16 bits and lose 4 again, the bot stops.
Maybe a sound notification would help to notify players their bot needs attention.

/just thinking out loud here

[KingZee]

I'd love to see more bot options, but I'm afraid it'll quickly become too complicated for new users. Therefore it's probably better to wait until there are much more active users before adding new bot-features.
It would be nice if bot settings can depend on the number of players left: if for instance you're 1 out of only 2 players left, the best strategy is to drown the other player all the time. However, he'll do the same, so you won't win much by adding this customization.

I would suggest to add a "take profit" and "stop loss" to the current "max loss". That way, you could for instance set these options:
-max loss: 2 bits
-take profit: 32 bits
-stop loss: 4 bits
So if you instantly lose 2 bits, the bot stops. If you reach 32 bits profit, the bot stops. And if you reach for instance 16 bits and lose 4 again, the bot stops.
Maybe a sound notification would help to notify players their bot needs attention.

/just thinking out loud here

I'm still emailing the owner, suggested for him to run a sig campaign, or a promotion/competition to reward the best bots of every category and promote that competition here, I can think of a million ways to raise an audience. So yeah, maybe it's best to keep it low on bot options for now, attract clients first.

And hey! Number of players was also going to be my 2nd suggestion after the owner implements multi-percentages. But actually, when I thought about it, fuzzy bots kind of have a point which is inacurracy. If given every variable, I am 100% confident in my skills to make the perfect bot. Of course that will make other people counter it, and then for me to keep updating it, which is in itself very fun. But with a weak amount of variables and more probability, it IS harder to make a winner bot, but at the same time I guess it benefits the website owner if I can't manage to do that.

For the stop-loss (in my opinion), I would just like a simple option that'll let me sleep in peace and let the bot run for 10^5+ rounds. It's frustrating that I have to worry about it every few times a day to make sure no one else managed to surpass my strategy and cost me hundreds of bits in profit. More options are nice too I guess, maybe after he implements the base one. :p

[LoyceV]

I'm currently beating 3 other bots in room 20 Moves. The moment a new user joins, or one of them changes their bot, this can be totally different again. Challenge: come and beat me

[jademaxsuy]

I think this game is good to play to waste time. I suggest to add a mode to play with real playera and not bots to make it more challenging. Bots is hard also and challenging too so playing with real players are more fun while earning also. I hope that the owner can find my suggestion (Just a suggestion though).