Bots have become a reality of online poker on some of very popular networks. While rooms do try to get rid of these poker playing programs, being able to identify and adapt against bots can help you a lot in the meantime.
This article may come
as a surprise to some of you. Why should we care about playing vs bots when we play online in an environment which involves human players? While we may believe this to be the case, it's unfortunately not true. Certain networks are infested with poker bots. It's only recently in the news that a bot-ring has taken millions of dollars from the mid-high stakes PLO tables.
The fact is that the majority of players would not even recognize if they were playing against a bot online as opposed to a real player. If you know what to look for then it can be blindingly obvious that our opponent is not human. Another sad truth is that reporting such bots to the network operators may have limited success. They will in some cases flat-out deny the existence of bots on their network. The reason? Poker sites like bots. They play big volume and generate a ton of rake.
We don't want to suggest here for a minute that all poker sites allow bots. Many of the bigger sites value their integrity and will actively hunt down and remove bots even if it means generating less rake overall for the network. Unfortunately other big sites will cover over the existence of bots and refuse to take action. In some extreme cases moving an entire bot-ring would potentially half the traffic at a certain limit and have a noticeable affect on the networks' revenues. It's not too difficult to see why in such extreme cases a network would have a vested interested in covering over illegal activity.
Even in cases where bot-rings are found on a network with good integrity, the good ones will not always be detected, which is not through any fault of the site itself. For example the recent million dollar PLO bot-scam occurred on a site with high integrity that does everything in its power to remove bots. Once the accounts were discovered they were naturally frozen and funds where seized.
Identifying bots crushing the high stakes can be difficult. They are obviously extremely well designed bots if they are beating the high stakes player-pool. However, most bots operate at nowhere near this level. The majority of bots can be found at the lower limit games – they exploit players who will have no notion that they might be playing against bots. It's also possible to beat lower limits with a far less robust bot, most people simply would not have access to a bot that can make millions after all. Such bots are simply not going to be made available for distribution, it's more profitable for bot owners to keep such bots running for themselves rather than to sell.
Most bots operate at the lower limit games, exploiting players who will have no notion that they might be playing against bots.
The bots at lower limits are far more obvious, and once identified they can actually be a profitable source of income. On one network I was making about 30buyins every 5k hands exploiting a huge bot ring I found. My graph had a positive red line and positive blue line. This is not the only bot-ring I have actively exploited and have found such setups on several networks. But how can we spot these bots in the first place? Here were the things that gave away one of the most profitable bot-rings I discovered.
- Every member of the bot-ring used the same prefop sizings (or multiples thereof)
- Every member of the bot-ring had almost identical stats over a large sample
- The bot-ring had a really obvious seating script where every bot would sit out at the same time
- The bot-ring took the same decision time for each hand
- Most bot-ring members had a similar pattern to their screen-name
This in itself is enough info for me to realise something strange was going on but not enough information to make money. This is when the experimentation began. I generated the following log of my experimentation.
Case Study – Crushing a Bot Ring
1. I notice I am being 3bet a lot in late position battles and start throwing in light 4bets. The consistency with which several players are 3bet/folding (with exactly the same size 3bet)is alarming. Easy game.
3. I notice that my 3bets are being 4bet to exactly the same size by the same group of players (round about 2.5x). I'm sitting deep so I begin 5bet bluffing small. Once again the consistency with which people are 4bet/folding in late position battles is alarming. In my experience 4bet bluffs are rare at the lower limits, and I've never seen such relentless 4bet/folding especially not to exactly the same size with 100% frequency by quite a large group of players.
3. It becomes apparent that these players have an extremely tight 4bet/call range presumably looking something like KK+, AK. I get glimpses of 3bet-bluff range which includes hands like Q4s and presumably other suited hi-cards. Bots seem to generally be firing off a cbet bluff in 3bet pots and giving up on hands below a certain strength. In single raised pots bots seem to be sometimes taking check/call lines on flop or turn with made hands such as TPGK.
4. I commence relentlessly 3betting/4betting every single bot I can find with any 2. The light 3/4/bets happen any time 2 of BU,SB,BB (i.e late position battles) are involved. Bot 4bets from CO,UTG,MP always seem to be value range that is getting reshipped so I 3bet/fold vs these positions. Mmm, profit.
5. I notice that my BU opens are no longer being 3-bet so light. What irks me is that NONE of the bots seem to be 3betting me light all of a sudden, not just the ones with history vs me. I notice my CO opens are now being 3bet light instead. I begin 4betting relentlessly vs these and making profit.
As for my BU opens it seems they are now only being 3bet with premiums. Bot players begin rarely 3betting and then 5bet shoving with a very high %. As you can imagine 4betting with 100% frequency vs a premium 3bet-value range is suicide. (Another telltale pattern these bots have; even when sitting over 200bb deep, 5bet is ALWAYS an allin)
6. At least....I assumed I was only being 5bet shoved on with premiums. Earlier today I caught a glimpse of that 5bet shoving range and now it apparently includes things like 99 and AQ. The complete range could easily be something like 77+,AJo+. Yet I'm convinced that initially I was only being 5bet with KK,AA so some serious adjusting has been taking place. The really suspicious thing seems to be that every single player I suspect of being a bot is adjusting in the same way at the same time. I might not play against one player at all who I presume to be a bot (name follows pattern); yet rather than that bot starting at level 1 vs me, it immediately starts playing vs me with a wide 5bet range.
By this stage I knew I was on to something. Evidently these bots were adjusting against me and had been programmed to do so based on my stats. I printed money for a while, then the bot-ring would adjust. I eventually discovered a way to completely destroy this bot-ring. Here is what I discovered.
- Bot ring was not sensitive to sizing. Facing a 3bb open, if I 3bet to 5bb I would get exactly the same amount of folds as if I made a 12bb 3bet. Hugely exploitable. My value range will now 3bet large and my entire 100% air bluff range will 3bet for a min-raise.
- Bot ring was check/raising aggressively on flop and had not been properly designed to counter flop 3bet aggression. Unlike the preflop scenarios where bot was adjusting based on my stats (known as a profile bot) the bot was not adjusting vs my flop 3bets. The bot was only continuing with the nuts in this situation.
- Step 3 was when I really started crushing. I noticed the bots had two different check/raise sizings based on whether they were value check/raising or bluff check/raising (facepalm). So every time they check/raise the flop large they have the nuts and I fold. Every time they check/raise the flop to a smaller sizing I insta click-it-back for 100% fold-equity. I was playing any 2 cards against these bots and looking for opportunities to exploit this and a couple of other serious massive leaks that I had discovered.
From this point onwards I had a consistent winrate of about 60bb/100 hands. I did this for several days and took a decent chunk of money from the bot-ring. Then the most surprising and hilarious thing happened which 100% proved that these guys were bots (even though it was pretty much proven by this point anyway). Every time I joined a table, every single player on my list of bots would immediately sit out.
They refused to give me any action whatsoever. In other words the bot owners had realised that they'd taken a huge hit over the last few days and had evidently programmed a killswitch to shut down all action against players that were exploiting them. I created an account on a different skin and continued crushing for another 24 hours before the killswitch was activated again.
Ethics and Denial
As far as I was aware I'd discovered conclusive proof that I was playing against bots. But could I really prove that these guys were bots? Prove is a strong word, and something that requires a huge sample size of hands. Proving someone is a bot is essentially impossible until we have an infinite sample. But if I can demonstrate beyond reasonable doubt that someone is a bot, should that not be sufficient evidence to categorically state that someone is a bot and ban them from the network? I could essentially predict which actions these guys would take and exactly when they'd leave the table.
So what did the poker network do when faced with such evidence. Nothing, they denied the bots exist. Even if the bots exist there is not enough evidence to prove anything. An interesting thing I noticed about the bots is that they clearly have an element of randomness programmed in – presumably as some type of primitive anti-bot-detection precaution. It involved donk-betting
. Each of these bots had a donk-bet flop of around 1%. It would rarely happen but every now and again each member of the bot-ring would randomly donk bet the flop (for exactly the same sizing). It was an interesting quirk and could potentially have been enough to throw me off the scent; that is, if every other bot-ring-member weren't doing exactly the same thing seemingly at random times.
Admittedly this could be enough to put off some type of bot-detection algorithm that the security departments of these various networks use. This is the only thing I can think of to say in their defense – whichever algorithms they used did not trigger a match as a result of this random anomaly that had been programmed into the bot. But it's a sad case if I as a regular player without sophisticated tools can spot something that an entire security team is managing to overlook.
This is of course not
what I truly believe. I simply believe the network did not care that bots were crushing recreational players at the tables. They realised that bots were contributing nicely to the overall rake generated and so were prepared to overlook something obvious. But is there something inherently bad about allowing bots to play at the tables? I would play against this bot ring any day of the week, perhaps it was even in my interest to not let people know about this bot-ring which was generating a nice source of income for me.
But the ethical issue underlying this is not whether bots should be allowed at a table. The issue is that pretty much all poker networks state in their terms and conditions that use of automated software to play their cash games is against the site rules. Bots could even be a good thing for poker – but the site needs to state clearly in their terms and conditions that bots are permissible at the tables.
The sad thing is that recreational players will not pick up on the same things that I noticed. I realised pretty quickly I was playing against bots and used it to my advantage. However, contrary to popular belief, some of these bots are absolutely crushing. I had to play a good 5k hands before I spotted major leaks
, and initially these bots were some of the toughest players to play against on the table. So for an average recreational player it's far better that the rest of the table is also recreational players rather than bots. They will have a much higher winrate this way even if it's negative overall. It's not fair for a bot to take money from a recreational player when the recreational player is told that there are no bots on the network. Anyway, now I'm telling you, watch out for bots at the lower limit games, they are there.
There is a darker and more sinister possibility here. If you had a ring of bots running on a certain poker network, would you have them all working independently or would you use information that one bot gains to help another bot in your bot-ring?
Sharing hole-card info could potentially allow bots to make better decisions when faced with all-in situations. They would know how likely it is for opponent to either hold or hit certain cards based on what their fellow bots had folded preflop.
As you can see in my note log of exploiting this bot-ring, certain bots were making pre-emptive adjustments based on information that they had not yet received. In other words if one bot in the bot-ring felt I was folding too much to 6bets, every other bot in the bot-ring would start 6-betting me with a wide range. In other words it's not just the no-bot policy that is being violated here. Direct cheating in the form of collaborating statistical info was happening.
And why stop there? If you are going to share a stat profile on a certain player why not share hole-card information with your fellow bot-ring players also? While it's not a good idea for me to share information on specific sites (don't ask me in the comments), there is one particular network which is notorious for it's RNG failing independent audits. Does this mean the RNG is rigged? Potentially, but there is another possibility.
Sharing hole-card info could potentially allow bots to make better decisions when faced with all-in situations. They would know how likely it is for opponent to either hold or hit certain cards based on what their fellow bots had folded preflop. It would appear to an outsider that the bot seemed to win more frequently than they should in all-in situations making a non-rigged RNG appear rigged over a large sample. For such an effect to be noticeable then a bot ring would need to very large.
If you spot a bot at the table there is a reasonable chance your poker network will care so let them know. We aren't trying to make all poker sites appear shady here. But assuming you find a bot and your network doesn't care, then use the above pointers to help you find ways of beating the bots. The first and most important step is identifying that one of your opponents is not human – so long as the bot is not super strong it should be downhill from there, experimenting until we find a way to print money.