Honest points, and why we do not say the games train your brain
· 4 min read · Tyamka team
How points work in all 15 Tyamka games, why results compare you only with yourself, and why we do not claim the games improve memory or attention.
Brain games have a habit of promising more than a game can deliver. We decided early that Tyamka would not. Below is where that line sits, how points are counted, and what the game does when your best cannot be loaded.
What we do not claim
In 2016 the US Federal Trade Commission settled with Lumos Labs, the maker of Lumosity, for 2 million dollars. The complaint was about advertising that claimed the games improved performance at work and school and helped against dementia and ADHD. A large review by Simons and colleagues, also from 2016, found that brain training mostly improves the trained tasks themselves. Evidence that it carries over to everyday life was much weaker.
We have no studies of our own, so we make no claim. On our game pages the question "Will it improve my memory?" gets the answer "We make no such claim. The points show how you did in this game."
In practice that rules out a set of words and numbers.
- No IQ, no brain age, no cognitive index.
- No percentiles and no comparison with the population. You will not see "better than 87% of players".
- No "improves memory, focus or productivity" at work, at school or in sport. We avoid even "trains" when it means a skill outside the game.
- No medicine at all. That covers dementia, ADHD and "protects your brain".
- No "scientifically proven" and no "designed by neuroscientists".
What we do say is simpler. These are short games for attention, memory and logic, and you can beat your own best. Where a game is a known psychology task, like the Stroop test or the Schulte table, we name it and describe what the player does.
One formula for points
All fifteen games use the same rule for faster play earning more. A correct action earns 10 points. On top of that comes a speed bonus of up to 10, which drops one point per fixed step of time. Each game sets its own step in its spec. The word game adds one thing, 5 more points for each letter over the minimum word length.
In the Schulte table the step is 300 ms, counted from the previous correct tap, or from the start for the first number. Find the next number in under 150 ms and you get the full 10. From 2.85 s on the bonus is zero, but the 10 base points stay. A slow, careful game still scores. A fast one scores more.
Every game's logic is a plain TypeScript module with test vectors, fixed inputs with the expected score. For the Schulte table part of those vectors came from an independent Python port of the spec, not from our own code. When the two disagree, one of them has a bug.
You against your own best
The results screen compares you with one person, yourself on the same level. It says what happened in plain words, for example "Level 15 complete. 43 of 49 found. New best, 58 more than before." It does not say GAME OVER, and it has no brain age.
We do not publish norms for our levels either. The difficulty tables are our own. Telling you what a good time on level 9 is would mean inventing a number.
A result that failed to load is shown as failed
If your best could not be loaded, the results screen says so. It does not show a zero, which would look like a real result. If a save fails, you see that too. Sync for signed-in players is written and still in testing. In it a smaller result never overwrites a larger one, on the device or on the server.
Fair timing in the reaction game
The reaction time test needs extra care, because timing is the whole game.
- The wait before the signal is random, from 1 to 4 seconds in whole milliseconds. It is not tied to the clock tick, so you cannot learn the rhythm.
- A press earlier than 120 ms after the signal is a false start. People do not react to a visual signal that fast, so such a press is a guess. A false start costs a life. If it only earned nothing, tapping every 150 ms would sometimes land on the signal by chance.
- On the web the press delay is zero for every button. React Native Web would otherwise drop very short taps.
What we cannot fix is the device. A screen, touch sensor or mouse adds its own delay before the game sees your tap, and that differs between a phone and a computer. So compare your reaction times on the same device.
All of this is also why leaderboards were hard for us. A public ranking compares players with each other, so it cannot simply take the score a stranger sends in. How we handle that is the subject of the next entry.