Dear readers, today we have a guest blog entry. Long-time reader Michael Peters likes poker and grew from a teenage punk rocker into a nerdy adult software engineer. He writes about Thomas Bayes, an 18th century English statistician, philosopher and Presbyterian minister. Reverend Bayes is best known today by math nerds. Like Michael.
In Michael’s words:
Back in the 1700’s Thomas Bayes came up with a statistical theory which today is used in all kinds of applications from pure statistics, to medicine and computer science. Bayes’ theorem provides a way to calculate the “probability of an event, based on prior knowledge of conditions that might be related to the event.” This basically means that if you know something about the chances that some characteristic predicts an outcome, you can then adjust the chances of that outcome for a specific event. You can also create a feedback loop so that the ultimate outcome of that specific event can improve your initial priors.
If this sounds confusing, lets try a more concrete example using something we all know and hate: spam email. Spammers are not sophisticated; they are just trying to send as much as possible so that they can get lucky a small percentage of the time. To catch it, all email filters have some base rules about what they consider spam. The more sophisticated ones learn new things based on their users’ preferences as the system sees more email (we’ll ignore discussions of Machine Learning and Artificial Intelligence and just focus on the statistical models). For instance, let’s say our filter has a rule that if an email contains the word “Viagra” (or variations on it like “v1agrA”) it has a 90% likelihood to be spam (the “prior”). For most people, most of the time this would be true. But if I’m ever in the market for little blue pills on the black market and they keep going to my spam folder, I’m likely to mark them in my email client as “Not Spam.” The more that I do this, the less weight the system is going to give to this prior and the less likely these emails will be considered spam in the future.
We don’t need to go into the actual formulas used for Bayesian Inference as you likely won’t be doing the calculations in your head. But it’s an important idea to have as you are trying to make decisions in a world of incomplete information. Start with assumptions you have learned over time (your “priors”) and adjust them as you gather more information about a specific situation. If you ever learn the “truth” about that specific situation you can then re-adjust your prior to better reflect reality.
Now let’s apply this idea to poker, the ultimate game of incomplete information. If you are playing online you don’t get to see your opponent. You don’t know their age, gender, education, profession or attire. In real life you might use all of this information to create your priors about how this person will play. Old man with coffee and a newspaper? The Bayesian prior could be that he’s very tight and won’t fight over marginal situations. Young unshaved man in a hoodie and headphones? The prior could be that he’s overly aggressive and not likely to engage in table talk. None of this applies to your online opponents. If we can’t gather those physical attributes to create priors, what can we use?
Lately I’ve been trying to collect my thoughts about what can be learned about a player’s screen name. Because I’m an American my online experience has been very limited. It’s hard to have enough data to create a detailed breakdown, but a simple binary “good” or “not good” is doable. The following numbers are somewhat arbitrary and your experience may differ, but here’s my breakdown from my previous experiences… i.e., my priors
80% Not Good
If the name meets one of these criteria, it’s 80% likely they are not a good player:
- They have the word “poker” in their name. We get it, you like poker. You’re on a poker site. This applies to words like “holdem” as well. Examples: “holdem-steve” or “P0k3r-lawyer”
- References to sex or genitalia. This isn’t a value judgement; most adults engage in and enjoy sex. But when it becomes part of your identity, it usually shows immaturity and not taking things seriously. Examples: “SexyLoverDude” or “big_swingin_ballz”
- Drug or alcohol references. Again, not a value judgement but when it becomes part of your identity you’ve probably spent a fair amount of time having it affect your judgement about your 40bb opening range from middle-position. Examples: “big420FAN” or “whiskey-and-blow”
- Bragging about their poker prowess. Words like “genius,” “boss,” “playa”, etc. Examples: “Mast3r_Rais3r” or “CardGenius”
- They are the hunter, you are the prey. Similar to the one above, but more directed at how much better they are than other players. Examples: “d0nK-Hunter123” or “KiLlEr_of_wHaLes”
70% Not Good
If the name meets one of these criteria, it’s 70% likely they are not a good player:
- Using some poker specific lingo almost like they are trying too hard to show they aren’t a beginner. Words like “nuts,” “flop” and “river”. Bonus points if they misspell it. Examples: “riverRAT35” or “OnlyBetsNUTZ”
- References the best hand in poker: AA. It’s almost like they think their name can change the probability that they will get dealt that hand. Examples: “justAAfish” or “texAAs-rose”
- Complaining about their bad luck, or praying for good fortune. Examples: “never-lucky99” or “Lady88Luck”
60% Not Good
If the name meets one of these criteria, it’s 60% likely they are not a good player:
- Talks about a sports team. I’m not sure why this one seems to be true. Most people who have competitive hobbies like poker also enjoy sports. And there’s a decent cross-over between professional sports bettors and poker players. But it’s been fairly consistent in my experience. Examples: “bullzFan79” or “blue-devilz-4eva”
- Insulting other players. Examples: “URaFISHY” or “Y-u-call”
Because of our knowledge of Bayesian thinking, we now know that these aren’t hard and fast rules. But when a new player sits down at your virtual table maybe you can get a little information about how well they play from looking at their chosen username. And as you see them play more hands, make sure to adjust that prior or you aren’t really putting Bayesian theory into practice, you’re just making snap, biased judgements.
So what poker screen names indicate that you’re a good player? To be honest, I haven’t come up with any consistent criteria. Maybe it’s because good players are also creative, so the names they come up with are myriad. Or maybe it’s similar to our spam example from earlier: good players put some effort into how they approach the game and bad players are just trying to get lucky.