
Breaking a memory game: a whole detective story

This game haunted us for several months. The rules are very simple: you need to lay out cards with pictures and come up with a fairy tale. Then remember her.
A simple thing to remember, we thought. They played it a hundred times, and all our education just yells that it will be simple now.
Three months. Three fucking months passed before we almost learned how to win. Along the way, we had to pick up translations of “War and Peace” and generally learn a lot about the length of words in different languages. At the same time we got to the psycholinguists. And they decided for themselves a long-standing question about the length of the message in Russian and English.
Spoiler: Russian on average (in long sentences) is no longer than English. In general, all languages are approximately the same in speed of information transfer - at least, as we were told at the RAS. But they immediately said that it is impossible to quote them, because there is no research, and REN-TV goes somewhere nearby. Here, the release of the English version of Habr with translations came in handy, so we quickly found the equivalent database.
But let's get it right from the start.
This is how it looks
One of the players says:
“We flew somehow on the plane,” and puts out a card:

You have these three cards:

You can just shut up, you can continue one of them in the spirit of “And then the aliens abducted us,” but you You decide to abruptly change the plot and say:

“And the funds of the Lenin library were being transported!”
The next player specifies:
- In Ryazan!

Further, the plot can go in any direction, depending on what the players put:

“Because all the books have already been digitized, and the originals went to feed silkworms on the new farm.” Or: “Because our friend Zinovy married the autistic girl Elena, and she calmly felt only in a house with many books. He bought the unnecessary in the libraries, restored them and turned his estate into a bookstore, but once ... ” Or: “During take-off, the main interior lighting was turned off, and we suddenly saw that there was someone among the books ...” In general, anything. And you continue. If there is a suitable pattern.
Then you will need to turn the deck over and tell a dream tale from the very beginning.
What is the ambush?
In time. You have only 2 minutes to put out cards.
To win, you need to score 61 points.
When you tell a fairy tale, then for each correctly guessed card you get 2 points, for a miss - minus two points. If you do not remember, but the neighbor can tell you - 1 point.

In our example, at first you do not see anything and must remember the beginning - "We flew in an airplane." You open the plane, this is it. You have 2 points. Then the next player recalls: "We brought books from the library." 2 more points. Next: "The light went out, and someone was there." Minus 2 points for missing Ryazan.
That is, you need to make 31 cards without errors. A better 32-33. And there is less than 4 seconds on the card to tell something about it, and everyone remembers.
The final task (for us, not for children) is to take 61 points, that is, lay out at least 31 cards and then not really mess up on restoring the plot.
Attempts to break the game in the first tests
In the forehead to take 61 points did not work. It seems that not the most stupid people gathered, but still to rise above 45 is really difficult.
The game seemed at first glance a cute childish thing, but then it directly hooked. We all felt that we needed to win those damn points. More precisely, beat the game. The box brazenly lay in front of us and provoked with its silence.
It became clear that there was no purpose to lay out more cards in a limited time. There is a goal to lay out 31 cards, and then remember them exactly. 33 or 37 - there is no difference. Before that, we were in a hurry to spend more cards.
We sat down again. The amendment to the strategy did not help.
We begin to break.
First pass optimization:The obvious mistake is that players try to use their cards somehow when they consider them suitable. And in the game there is no relation to how many cards you discard. That is, this action is clearly dictated by some past experience or desire to participate. In a degenerate case, one player can tell a tale, and the rest should remember it. The point is this.
We agreed to avoid unnecessary use of maps so that the plot does not turn sharply towards the drawing. That is, so that there are no twists like: “And then they took and went to the mountains!”.
Did not help. More precisely, it helped, but not much.
We tested the hypothesis with one storyteller . It doesn’t roll, a set of 3 cards in the buffer nevertheless fits worse than a set of 12-15 cards at all at the table.
Another hypothesis: on the “promotion” of the story, everyone put cards in a pseudo-random order, but you need to tell one card in a circle. Is it possible to agree that we will lay out in the order that we tell, so that it is easier to remember?
We check the same pitchfork: the plot needs to be rotated too much to use the available pictures in the buffer of 3 player cards. Empirically establish that these costs of memorization are stronger than the gain from the story about their cards.
Another hypothesis: is it possible to designate one player as a memory player in order to contact him for 1-point cards in case of problems? The result is no clear benefit. Own participation helps to memorize pieces of the plot nearby.
The rules do not prohibit discussing the following cards. Players begin to tell not only their card, but also give tips on the next, if they remember. We feel that this breaks the spirit of the game, and still does not give a clear gain.
We continue to break hard.
The hypothesis of them. Dijkstra: it is necessary to remember pairs of cards on one plot block . That is, make up such "syllables" from cards. It doesn’t work directly, but it comes to understanding that after 20-30 games in the same composition, plot macros are formed from sets of cards. The question is:
- It smacks of cheat (we should not remember the plot of the last installment or do precalculations)
- It’s not always possible to collect the necessary cards for such a macro - in the game, not the whole deck at the same time.
Hypothesis: we are dumb or too self-confident . It should be given to another group for tests. We check. Cannot take more than 45 points.
We study the experience of other groups and the experience of the video authors of the game. Here the reader of our telegram channel @ mosigra helped us with the analysis , where we sometimes write about the development process. So, their course in successful parties is a phrase in 2-4 semantic units (“and then someone died”, “my wife and I danced at the wedding,” “it was a million years ago”). It’s not immediately obvious to our person that in a game of fantasy and storytelling, a move is the shortest possible phrase with as few words as possible. This is counterintuitive. For us, this also did not follow the rules, but we already guessed. True, they didn’t do it so short.
Later, we made a change to the rules about the method, and in the release we recommend saying one sentence, not several. But at that moment they tried to play again. As it turned out, ultra-short forms do not allow to build a vivid plot, that is, we run into poor memorability with a good number of cards. One sentence is quite normal.
After several weeks of training, we bring the indicator to 51, Max (development manager) once saw 55, but could not repeat it.
Just a couple of millimeters is missing.
Hypothesis: perhaps the English language is faster , and in Russian it takes more time for the words themselves. And the most valuable resource after the accuracy of the map is just the time for a story.
Because we know the story about the mat. By the way, they also tried the mat, it does not contribute to memorization.
A whole study was needed to test the hypothesis.
Testing the language hypothesis
Anya from the development contacts the psycholinguist of the Russian Academy of Sciences (the answer was not official, and she asked not to mention it, so there will be no more details). It gives her our hypothesis about the difference in the speed of English and Russian spoken language, adds a hypothesis about the difference in the semantic content of words in different languages. The lady suggested conducting an experiment with a voice recorder and phrases in different languages. The general answer is that he does not know any ready-made studies on this topic (namely, when it comes to comparing English and Russian languages), but the idea is this: there are various possibilities for linguistic coding, but that, regardless of whether the language is fast or slow, the efficiency of information transfer is They are the same. For example, fast Spanish consists of many syllables, each of which in itself does not mean anything or means little.
Plus there are extralinguistic factors, for example, individual characteristics of speech. Well, roughly speaking, some people have a fast pace, others have a slow pace, and this may not even be related to what language a person speaks.
In general, let the holivar about the length of the tongues begin!
In the meantime, we are returning to the experiment with the game.
The following hypothesis about psychology
We contact the author, we find out that it usually takes several weeks before the children reach their level. That is, they really train (but children under different ages have different points limits). Ok, great, we have a whole familiar test kindergarten, in which there are children with cognitive impairment, and just children. And professional psychologists. We give the game to see them (alpha box). Perhaps we learn something new from their analysis.
Feedback comes. I will quote the most important parts of the conclusions for the development team.
“For preschoolers, the game is good as a means of developing coherent speech. But since it’s a very difficult task for many of them to speak coherently, it’s better not to use the hourglass at all, so as not to make the difficult task even stressful ... it’s worth considering the issue of watches for 3-4 minutes ” - ok, great, it’s useful for children . This is good for the karma of the game, but still does not help us win.
“The optimal players when playing with preschoolers are three people, two of whom are children, and the third is an adult. Together (a child and an adult) is also good, but the three of them are more fun. If there are more players, the children have to wait too long for their turn to speak out and they begin to get bored. It’s not very interesting for them to listen to each other, it’s much more important for them to express themselves. ”- hmm, it’s clear that it is necessary to sell it to families, at small libraries, small children cannot themselves without an adult.
“The task of linking the next fragment in meaning with the previous ones is very nontrivial for preschoolers, often for younger schoolchildren too. By default, the child seeks to use a purely mechanical stringing of plot elements (looked up, saw this, then went there, that was there), so the story in the end most resembles a trip to the store. Here, it’s just very important that an adult participate in the game, who can demonstrate by personal example the formation of some more complex semantic connections. At this place, the development of thinking ” - take into account.
“The mood of the cards is very conducive to talk about how the characters in the story feel and how they react to current events. Well, in no hurry, discuss this at every opportunity. Here the development of emotional intelligence takes place (by the way, it’s rare where you can immediately develop both emotional intelligence and logical intelligence at the level of understanding cause-effect relationships) ” - this we will tell reporters, excellent.
“When playing with preschoolers with a sequence of moves, it makes sense to act not as written in the general rules, but exactly the opposite: compose in strict order, and remember in chorus. The turn of the course provides everyone with an equal opportunity to speak, and this is important because everyone wants to speak, but some children think and speak more actively and faster, and if you do not follow the sequence, they simply push those who are slower in temperament. On the other hand, if they remember everything together, then the child does not have the feeling that he is being interrogated. It makes sense to remember to victory, throwing leading questions until the team remembers everything until the last card, so that there remains a sense of success and completeness. For preschoolers in this game, the process itself is much more important than productivity ” - An excellent option for homerun, we will transfer this to the recommendations for the game directly on the product card.
That is, it does not help us to win, but explains that for children, the timer in life is not the main thing. And hide the cards too. But the study reveals another layer - the way children build associations with cards.
- The picture with the green elf ... provokes a violent reaction among preschoolers like “Oh, he has a naked priest”. For a preschooler, the view of naked buttocks is pornography, and the response provokes the corresponding. If the elf cannot be equipped with robes, then for playing with preschoolers it is better to simply remove it.
- A picture of a kissing bride and groom may confuse some children. Many preschoolers consider the kiss to be somewhat indecent and react to pictures with kisses again like pornography - with lively, but somewhat redundant interest in the game. The effect of the fact that in our culture it’s not very customary to kiss with children.
- A picture with a body outline drawn in chalk does not tell preschoolers anything and is not recognized by them (“this is some kind of boy”). Before prepubertal inclusive, this is a little relevant image.
- The call for help, written in the sand, is language-dependent, so it makes no sense to use the corresponding card when playing with a child who does not yet know how to read.
- The image of hippies is not recognized by preschoolers as hippies, because they do not even theoretically imagine what it is. Interpreted as "some kind of tough guy." This card is more likely for adults, because they have something to invest in it; children it is obscure.
- The image of punk is a little closer to the experience of children, because live punks are sometimes found here, but they are perceived by them again as something not quite decent. This card will become truly relevant only to puberty, when it is overgrown with some meanings.
- A card with the number "13" requires not only the ability to recognize numbers within two dozen (6-7 years), but also the understanding that this is an unlucky number. In general, this is a rather unobvious association for a child.

That is, they recognize punk, but hippies are gone. But still, psychologist’s tests do not help us win.
The answer comes to the hypothesis of the duration of the tongue
We are still doing our research.
At first Habr really helped. There are also technical texts with a good rating. If both versions of the text - in the Russian and English versions - have ratings not in the ass, then most likely either the translation is high-quality or the material is relevant. We copy the translations, we find their direct analogues in Russian. Let's look at modern ones (then we looked at a smaller number). In the URI field - the post number (it can be appended to the end of the link, for example, the first one about Go is, as it might seem, at habr.com/ru/company/mosigra/blog/451938 , but in fact in the Buda blog - there will be a forward ) Further words in two versions and signs in two versions (with and without spaces):
Headline | Post number in URI | Words in EN | Words in RU | Marks in EN | Marks in RU |
Go Bitmap Indexes: Wild Search | 451938 455608 | 3 698 | 3,548 | 21,449 17 937 | 24,428 21,065 |
We modify the Bluetooth stack to improve the sound on headphones without codecs AAC, aptX and LDAC | 456476 455316 | 1 921 | 1,676 | 11 398 9 588 | 11,494 9 945 |
Audio via Bluetooth: maximize details on profiles, codecs and devices | 456182 427997 | 7 009 | 6 061 | 41,723 35,212 | 42 959 37,426 |
The Baikal-T1 processor and the BFK 3.1 PAK: first tests | 440888 967191 | 2 723 | 3 034 | 16,246 13 577 | 19 477 16 921 |
Using signed bootloaders to bypass UEFI Secure Boot protection | 446238 446072 | 1,464 | 1,272 | 9 631 7 624 | 10 069 8 255 |
How Protonmail is blocked in Russia | 443638 443222 | 2 418 | 2,314 | 13,996 11 761 | 15 194 13 066 |
I spoil the code of my life for developers and I don’t want to do it anymore | 440736 432822 | 1,582 | 1,415 | 8 620 7 069 | 8732 7,339 |
Quite fanciful "Beginnings" of Euclid in TeX | 452520 451682 | 2 002 | 1,758 | 11 730 9 651 | 11 838 10 002 |
“I can tell you about the common pain of all iOS developers” - 10 questions for the programmer, issue 2 | 438512 420981 | 2,260 | 1,904 | 12 512 10 297 | 11,890 10,026 |
Homemade text laser projector | 438618 407591 | 1 929 | 1,785 | 11 206 9 344 | 13,239 11 523 |
I lost faith in development, burned out, but the tool cult saved me | 440914 436202 | 2 011 | 1 794 | 10 876 8 901 | 11 065 9,292 |
How to milk cows with robots and make an industrial startup on it. R-SEPT Development History | 436000 430740 | 3,214 | 2,535 | 17,872 14,710 | 17 178 14,689 |
“I am a useless fool and want to quit” - 10 questions for a programmer, pilot release | 436012 420321 | 2,035 | 1,557 | 11,233 9,259 | 9,820 8 319 |
Total English versions:
• 34,266 words.
• Characters with spaces - 198,492.
• Characters without spaces - 164,930.
Russian versions:
• 30,653 words.
• Characters with spaces - 207 383.
• Characters without spaces - 177 868.
The difference is that Russian is shorter by 10.54% in the number of words, 4.29% longer in the number of characters and longer in the number of characters longer by 7.84%.
That is, judging by Habr, we can add 4 points to us. Hooray, that's already 59 out of 60 needed!
But! These are technical texts, right?
Translations are not two-way. Translation of the translation does not lead to the original. If anyone is interested in being a geek, start with a great Mogwaika post . Very briefly: there is meaning and there is content. Alekseeva I.S. in "Introduction to Translation Studies" gives an example:
“Russian phrase:“ Citizens, do not forget to pay the fare! ” - identical in function to the German: “Wer ist (noch) zugestiegen?” (Who else came in?) And encourages the recipients to do the same actions, although the objective situation (the situation described in the text) and the not-so-identical speech situation (the situation in which communication takes place) seem to overlap, forming a situational context. ”
In general, you can translate in different ways. If the focus is on normative content , you get a translation of the contract when it is important not to miss the transaction, and everything else burns with a blue flame. Of the works of art, it is ideally suited only for the genre of “police report”.
Focus on context- translation of poetry. This is when the meaning of a phrase becomes more important than the meaning of words. An example of interest: “author of a board game” and “game designer”. The first phrase in English simply does not make sense, because the author is not at the table, but at a literary work. And a game designer means “developer”, to which the population is not yet accustomed. And he draws nothing. Therefore, it is important to translate here not in a normative-meaningful way, but in a contextual way. Another example of translation in context is, for example, time spans across centuries and cultures. The point here is not even the reduction to SI, but the fact that the Mongols have a “hour” (actually of a different length) tied to the place of the sun relative to the entrance to the yurt, that is, the hour is shorter in winter and longer in summer. If anyone is interested in this offtopic, here are the details already in my channel.
The next criterion is called "full translation" - this is when the content is transmitted by equivalent means. A more advanced level is when a person does not just receive the same information, but is formed in such a way that it causes the same cognitive reactions. Actually, a comparison of the reactions of the recipients is the concept of dynamic equivalence. A side effect - if you are not very skillful in using such a concept, then you will get the villain of Evil.
In general, we need dynamically equivalent translations in both directions, so this is such a complexity.
Let's look at foreign literature. And here Flibust will help us) We need a good literary translation, because it conveys meaning as is. We find books in English, compare 5 different pieces with a good Russian translation - also in the number of words and characters with or without spaces. We write down the names of books and translators. Then we find classics in Russian which were translated into English. We take 5 different books by different authors. We write down the names of books and translators. Compare the length in the same way.
Оригинал 3 главы — 2827 слов, 11544 символа без пробелов, 14317 символов с пробелами.
Перевод 3 главы — 2208 слов, 11579 символов без пробелов, 14568 символов с пробелами.
Эрнест Хемингуэй — Старик и море (перевод Е. Голышевой и Б. Изакова)
Оригинал — 26599 слов, 106064 символа без пробелов, 110946 символов с пробелами.
Перевод — 23175 слов, 134502 символа без пробелов, 135917 символов с пробелами.
Джон Стейнбек — Гроздья Гнева (перевод Н. Волжиной)
Оригинал — 179691 слово, 773589 символов без пробелов, 947167 символов с пробелами.
Перевод — 151475 слов, 767191 символ без пробелов, 971718 символов с пробелами.
Филип К. Дик — Мечтают ли андроиды об электроовцах (перевод М. Пчелинцева)
Оригинал — 61953 слова, 287196 символов без пробелов, 346956 символов с пробелами.
Перевод — 59520 слов, 329910 символов без пробелов, 391435 символов с пробелами.
Сомерсет Уильям Моэм — Бремя страстей человеческих (перевод Е. Голышевой, Б. Изакова)
Оригинал — 259601 слово, 1143692 символа без пробелов, 1397103 символа с пробелами.
Перевод — 211057 слов, 1128656 символов без пробелов, 1340885 символов с пробелами.
• According to foreign versions longer by about 15%.
• The signs are approximately the same (2% and 1% shorter).
That is, the translation of the English version into Russian gives about the same message length. Now let's look at the reverse translation - attempts to express Pushkin through someone else's cultural code.
Оригинал — 5605 слов, 28804 символа без пробелов, 34460 символов с пробелами.
Перевод — 6957 слов, 30870 символов без пробелов, 37701 символ с пробелами.
М.А. Булгаков — Мастер и Маргарита (перевод Л. Волохонской)
Оригинал — 112149 слов, 629215 символов без пробелов, 742796 символов с пробелами.
Перевод — 145291 слов, 692753 символов без пробелов, 834321 символ с пробелами.
Ф. М. Достоевский — Преступление и наказание (перевод Л. Волохонской)
Оригинал — 169908 слов, 895178 символов без пробелов, 1067426 символов с пробелами.
Перевод — 213371 слово, 990536 символов без пробелов, 1199117 символов с пробелами.
А. С. Пушкин — Евгений Онегин (перевод В. В. Набокова)
The original is 23046 words, 119022 characters without spaces, 136648 characters with spaces.
Translation - 32501 words, 147914 characters without spaces, 179655 characters with spaces.
B. L. Pasternak - Doctor Zhivago (translation by L. Volokhonskaya)
Original - 153093 words, 864243 characters without spaces, 1013754 characters with spaces.
Translation - 210573 words, 990388 characters without spaces, 1196558 characters with spaces.
• According to foreign versions longer by about 30%.
• By signs, foreign versions are longer by 12% and 15%.
That is, what we transfer from here to there becomes longer. Although, it would seem, you expect the opposite.
Total
The hypothesis is that the problem in us, as players, is most likely. It is assumed that the game will develop us, and it does. As a result, " Time to sleep " came out with constant thresholds for points.
As the skills of memorization grew, we saw progress, but we still did not play enough to win the game, stabilized somewhere around 54 out of 60. Our colleague scored 58 points out of 60, but no one saw this - however, the result is inspiring optimism. Now I suggest you try to do the same.