Your opponent tags

The previous chapter showed you the only categorization Poker Spin Tracker handles entirely for you: fish or reg. Tags are the next step: your own labels, placed on the opponents of your choice.

Tags are a very powerful tool, but also a very treacherous one. A rule-based tag rests on the individual stats of opponents you have sometimes faced for only ten hands. The grouped analyses that come out of it can then be completely wrong while looking perfectly solid.

So this chapter reads in two parts: first building and using your tags properly, then understanding, numbers in hand, the statistical biases they can manufacture.

There you tick the tags you want, and the New tag entry lets you create one on the fly, immediately assigned to the player.

A tag placed by hand is marked "manual" on the Players page, and it takes priority: automatic recalculations never remove it. You decide, the engine yields.

Assignments are recomputed at three moments:

• on every import, for the opponents present in the imported hands;
• when a rule is created or edited, for that tag across the whole database;
• when you click Recalculate stats on the Players page, which runs a full pass.

Everything is driven from the Players page, Tags tab (the switcher at the top of the page offers Players, Tags, Aliases and HUD). There you find a table with one row per tag:

The CEV columns deserve a second look: they don't show your results against these players, but the CEV of the tagged players themselves, in chips per tournament, computed on the hands you played against them.

It's your profitability barometer for a cohort: the more negative the tagged players' CEV, the more chips that profile bleeds, and the more profitable it is to face.

Three families of rows live side by side in this table:

• The default tags: Fish, Reg and Unknown, the ones from the previous chapter. They are locked: they can be neither renamed nor deleted, and their classification logic stays that of the automatic algorithm.
• The suggested tags: two ready-to-use profiles, Passive and Aggressive, marked with a "suggested" dot (detailed just below).
• Your custom tags, editable and deletable through the pencil and bin icons on each row.

The Passive fish is clearly more profitable to face than the average fish; the Aggressive fish, on the other hand, is clearly less so. Knowing which of the two sits across from you genuinely changes the way you approach the game.

The creation panel asks you for a name, a color, and above all conditions, which you add one by one and which stack up as cards. Five tabs organize the work:

The first three tabs set conditions; the last two set nothing, they narrow the measurement window of every condition at once.

On top of that, each condition carries its own scope, adjustable in the right-hand column:

• Format: 3-max, Heads-up, or both;
• Positions: BTN, SB, BB, and their variants against each opponent;
• Pot type (for postflop stats): raise pot, limp pot, ISO pot, 3-bet pot;
• Vs: measure the stat against everyone, against you only, or against the players carrying another tag.

Creating a tag isn't enough: it only becomes truly useful once you use it to group players together and filter your analyses.

This is the filter that answers composition questions:

Or again:

The dashboard's CEV by profile view rests on the same logic, and we'll come back to it in its own chapter.

Computed tags rest on the individual stats of each opponent, measured only on the hands you played against them. In Spin, that sample is often tiny: about ten hands for most players.

And there is something nastier: regression to the mean

Take a "VPIP ≥ 55%" tag computed over about ten hands. Who falls into it?

The genuinely loose players, of course. But also every ordinary player who simply got good cards during those ten hands. And since ordinary players are far more numerous in the pool, they end up in the majority inside your cohort.

The other way round, the truly loose player who was dealt ten unplayable hands slips through the filter and never makes it in.

The result: your "VPIP ≥ 55%" cohort actually plays clearly fewer than 55% of hands. The threshold you set is not the group's real level, it is the crest of a wave of chance.

An adjustment calibrated on the displayed number will therefore overshoot by a long way compared with what these players really do.

Here now is the most insidious trap of tags, the one that on its own justifies this second half of the chapter. Picture it: you enable the Aggressive tag, then you filter your dashboard on the spins where an opponent carries that tag. And you discover:

No. Every tracker user sees the same thing with this filter, whatever their real luck. It is neither a CEV bug nor a filter problem: both curves come out of exactly the same hands. It is a selection bias, and it is worth understanding step by step.

The mechanism, in three steps

1. When does an opponent show a high VPIP, PFR or open shove % before your eyes? Mostly in heads-up, where everyone plays far wider, and when stacks are short, in push/fold mode. Their observed stats explode during those phases.
2. But you only reach the HU of a Spin if you survived the 3-max, very often by winning your flip. The long HU phases, the ones that inflate your opponents' aggressive stats, belong to the tournaments where you have already piled up chips.
3. Since the tag is computed on those very hands, filtering "vs players observed as aggressive" largely amounts to selecting the tournaments where you won your all-ins. The chips curve lifts off from CEV by construction.

We measured this gap between chips won and CEV, tournament by tournament, on a test database of several tens of thousands of spins:

And here is the proof that the culprit really is the small sample: the bias is massive when the tag rests on few hands, and it disappears completely when the tagged opponents have a real history.

The same mechanism plays out the other way round for passive profiles. An opponent looks passive to you mostly when you see them play little, so during the deep-stacked phases of the 3-max: the ones where the three stacks are still close and nobody has taken the lead.

Filtering "vs passives" therefore amounts to selecting the tournaments where you never took control of the chips, often the ones where you busted early. This time the chips curve runs below CEV: fake bad luck.

This trap is only one example. The mechanism that produces it (stats measured player by player on few hands, then read at the scale of a group) manufactures several others, subtler ones. Here they are.

Three more traps, worth knowing before any cohort analysis.

Everything above condenses into a short list of reflexes:

1. Demand a sample, on both fronts. The two settings do not measure the same thing (see just below): raise the "over ≥ N opportunities" of every condition to 8 at the very least, like the suggested tags, and add a Volume ≥ 50 hands condition as soon as the tag is going to serve a fine-grained read.
2. Prefer stable stats to game-phase stats. Limp and PFR describe a player. The open shove or all-in % mostly describes their stack at that moment. If you use a phase stat, bound the effective stack and the position so you compare like with like.
3. Keep it lean: two or three conditions at most. A simple rule can be understood, checked and corrected. A rule with eight conditions is a tag you will no longer be able to read back in a month.
4. Start from the two suggested tags. Passive and Aggressive are calibrated on a large volume: enable them, watch them, and only then create your own variants.

Let's sum up by separating three uses, because the answer is not the same for each of them.

For the big pool reads, the aggregated vs Fish stats remain your most solid reference: the cohort is huge, stable, and your exploitative ranges are already built against it.