We celebrate a streak that a coin gets 81 percent of the time

Get two answers right in SwipeTA and a small chip appears with a flame on it. Keep going and it counts up. It is a good feeling, and it is meant to be.

We have also published the arithmetic showing that over a hundred answers, a player choosing completely at random will hit a five-in-a-row streak 81% of the time.

So we are celebrating something a player with no skill at all will produce anyway. This is about why we do that on purpose.

The game part is real, and we are not embarrassed about it#

It would be easy to describe SwipeTA as an austere measurement instrument that happens to have a UI. That is not what it is.

Getting one right slams a CORRECT stamp onto the screen at nearly twice its final size, glowing, before it drifts off. There is a mascot, T.A, who turns up in the draws, on the home screen and in the results. There are six board skins. The rank ladder is eight tiers of six steps — 48 in all — and promotion lights one of them. Finishing rounds earns draws for a collectible cast. There is a training journal with its own charts.

Every screen you will see several hundred times has been treated as something worth making properly. That is a deliberate allocation of effort, not decoration left over from a design phase.

None of it makes a chart answer more informative. Its job is to make another answer happen.

What the streak counter is actually for#

Here is the tension stated plainly.

In a language app, a five-in-a-row streak means you knew five things. Here, over a hundred answers, 81% of players who know nothing at all will hit one, and 32% will hit seven in a row. A single twenty-answer session scores 60% or better a quarter of the time on luck alone. We worked all of that out and published it.

So the chip is not evidence. What is it?

The answer is in a number from the same calculation. Establishing that a genuinely 60%-accurate player really is 60% accurate takes about 194 answers. A 55% player needs 783. Those are not numbers you reach in a session, or in a week of sessions you had to talk yourself into.

The celebration is not a claim about the answer. It is what makes the sample size happen.

That is a less flattering description of game design than most products would give, and it is the accurate one. The flame is not telling you that you read the chart well. It is making it more likely that you are here again tomorrow, which is the only route to a number that means anything.

Celebration and measurement are different systems#

It is worth naming the split, because most of the design follows from it.

The streak counter is motivational feedback. The statistics panel is an inference system. We deliberately do not ask them to do the same job. One of them is allowed to be generous, because being generous is its entire function. The other is not allowed to say anything until the sample supports it.

Confusing the two is how a practice tool ends up telling people they have an edge because they had a good Tuesday. Keeping them apart is why the flame can start at two answers while the read on you stays shut until 85.

We argued the first half of this in bar replay is a better chart, a question bank is a better test: controls make a single answer meaningful, but only volume makes a set of them readable. This is the other half — how the volume gets there.

Where we refuse to gamify, and why that is the clearer statement#

Anyone can list the rewards they added. What a practice tool believes shows up in the places it declines to reward you, so here are ours.

The statistics stay locked until answer 85. The app will not tell you that you lean long, or how you do after a wrong answer, until it has seen enough answers for that to be a finding rather than a plausible-looking accident. Eighty-five is where the arithmetic lands for detecting a 65/35 lean, and the panel shows a countdown instead of a number. A number at answer 20 would be more engaging and would be noise.

XP pays only the first time you get a question right. Classic levels are fixed sets, replayable on purpose — re-clearing a board you know is how confidence gets built. But a replay earns nothing, and cannot be used to farm the draw meter. The reward is attached to clearing new material, not to replaying known boards for currency — and note that we are careful not to call a first correct answer "learning". It might have been a guess. That is the whole problem this article started with.

Draws are earned in play rather than sold, and the odds are published. The collection is cosmetic and has no effect on the training. You can read the drop rates before you decide to care about them.

The correct-answer stamp is tinted by board skin — deliberately, because it marks correctness and not direction. It is a small thing that says the whole architecture out loud: a player can recolour the celebration freely, because the celebration is not carrying information about up or down. What the reader is allowed to restyle and what they are not is a map of what we think is load-bearing.

The part we are least comfortable with#

A tool that celebrates also teaches you to expect celebration, and there is a version of that which is not good for anybody.

We have some protection from the worst of it — the questions are balanced so you cannot farm a score, there is no leaderboard for Classic, and there is nothing to buy that improves your results. But the honest position is that we are using motivation techniques on a task where the feedback itself is unreliable, and we do not know how to separate "you enjoyed that" from "you learned something" cleanly here.

What we can do is refuse to let the enjoyable parts make claims. The flame counts answers; it is not allowed to draw a conclusion from them. That separation is the one part of this we are confident about.

What none of this claims#

Not that celebration makes you better at reading charts. Not that streaks mean anything about your next call. And not that any of it transfers to a real market — the transfer question is open, no study we know of settles it, and we said the same thing in what the research says about trader intuition.

SwipeTA is a training game and a simulation: no real money, no broker, and it does not provide investment advice.

The claim is narrower and, we think, more interesting. Volume is the only thing that turns a practice session into evidence, and volume is a design problem before it is a statistical one. We built the fun part because the maths demanded it, published the maths that says the fun part is not proof, and kept them apart. The parameters behind our own numbers are on the methodology page.