2018-06-01

Durak Win Rate Analysis

Analyzing the win rate for strategies in the Russian game Durak (Fool)

As a fond player of the Russian game Durak, I wanted to test how much of the game is Strategy versus Chance.
I paired a "Beginner" and an "Expert" against each other in a series of simulated millions of games.
The Beginner simply attacked and defended with the smallest playable card.
The Expert used every single strategy possible to win.
After much trial and error, I found strategies that increased the chances of winning, and ones that didn't.
This game was "Reversable Durak", meaning the defender can reverse the cards back at the attacker, if the defender has a matching card.
Unlike most online Durak games, I made sure that the computer never "cheated".
The players were only provided the following information:
  • The players' own cards
  • The cards that are currently being played on the table
  • The Trump/Kozer
  • How many cards are left in the deck
The players had no ability to look into the opponent's cards, the deck's cards, or the beaten "out" deck.
The players also had no ability to remember old played cards.
I did this because human players don't have the ability to remember all old cards, so it's unfair for a computer to do it.


Check out the code and analysis here - https://github.com/veniamin-ilmer/durak-analysis

2018-05-16

Learning to Read and Write Chinese

I enjoy learning languages. Some examples of languages I played with were Spanish, German, Esperanto.. I even dabbed into Hebrew, and ancient Greek.

I've made a few observations about my language learning:
  • Synthetic languages, like German or Russian are harder for me to remember, due to the memorization of the long words and spelling involved.
  • I pick up how to recognize characters relatively quickly. For example, I learned how to recognize Hebrew script in about a day. Same with Greek letters.
These two points made me interested in Chinese.
Chinese is an Analytic language, and involves recognition of lots and lots of characters.
Instead of spelling, you're drawing a picture.

And so, two weeks ago I embarked on trying to learn Chinese... And quickly, I ran into problems.

The main reason I decide to try out Chinese, was because I think I can pick up reading characters quickly.
Yet, most of the time learning tools are built to value speaking more important than reading/writing.
The Chinese character is almost looked at as an inconvenience, due to it being difficult to type using Chinese characters online.

With focus on speaking more than reading/writing, a lot of words are taught as a combination of multiple Chinese characters, without breaking down what each character means.
For example, they may teach 名字 means "First Name", without explaining what does 名 or 字 mean.
And just to let you know, neither 名 nor 字 means "First" or "Given" or "Nick".

As someone trying to read and write each character, this makes it very difficult to learn.

And so, I came up with my own system to learn these.
Using various language learning tools, I build up a list of new vocabulary words per subject.
I then load up the vocabulary into quizlet.
I break up the subject into two sections: Learning the characters independently, and then combining the characters together to form new words.
Finally, and most importantly, I open the quizlet app on my phone.

Writing in Chinese on the phone is the best part.
You can configure your phone to write in Chinese... Not by typing, but by literally drawing each character with your finger.
Not only did writing in Chinese characters help me recognize them easily, but it also taught me about the correct stroke order that you're supposed to write in.
Through trial and error, I got good at it. And Android's AI got better at recognizing my hand writing.

I've made my study sets available to the public. They can be accessed here:
https://quizlet.com/litehacker
The desktop version of quizlet is a little slow/bugging, but the phone app version is amazing. I highly recommend opening it on the phone and configuring the chinese language input to be in written form.

Start with Unit 1, and keep going up in numbers. For each unit "A" contains the individual characters, while units with "B" contains a combination of those characters. This makes the transition to learning more complicated character combinations much more simple.

I'm in the process of adding more and more vocabulary.
The most difficult problem with Chinese is that frequently words are taught without breaking down the characters.
Most of my time and effort is spending time deciphering what each character individually means.
Once I have done the research, I add them to the list.

I hope in the future to be able to read Chinese websites, and find out what that closed off side of the world is all about.

2018-03-09

Study Review: Linoleate-Rich High-Fat Diet Decreases Mortality in Hypertensive Heart Failure Rats Compared With Lard and Low-Fat Diets

Study: http://hyper.ahajournals.org/content/hypertensionaha/52/3/549.full.pdf

Energy Macro Percent Usage
DietCarbsFatsProtein
Lard43.5%42.2%14.3%
HLSO43.5%42.2%14.3%

Historic charts show rats fed HLSO had a longer lifespan than rats fed lard.

CriteriaScoreComments
One Constant Macro****Article notes both diets to be exactly the same, but does not detail formulation.
Constant Nutrients + Fiber***No breakdown of food provided. Article states that all diets were based on the "Purina 5001", but purina's official datasheet differs.
Number of animals per diet****13, 13, and 16 rats used.
Length of Study*****The whole length of the rat's life.
Funding Bias*****Grants from American Heart Association and NIH
Total4.2Good Quality Study

2018-03-06

Study Review: Life-Span Extension in Mice by Preweaning Food Restriction and by Methionine Restriction in Middle Age

Study: http://www.pkdiet.com/pdf/LifeExtMice.pdf

Energy Macro Percent Usage
DietCarbsFatsProtein
0.43% Methionine69.6%8.1%12.5%
0.15% Methionine69.9%8.1%12.3%

Historic charts show Methionine restricted mice lived longer.

CriteriaScoreComments
One Constant Macro*****Diets were almost identical, with exception of the Methionine.
Constant Nutrients + Fiber*****Diets were almost identical, with exception of the Methionine.
Number of animals per diet*****43 mice were used per diet.
Length of Study*****1200+ days. The whole length of the mouse life.
Funding Bias*****NIA grants and American Federation for Aging Research Grant
Total5.0Good Quality Study

2018-03-03

Study Review: Fructose decreases physical activity and increases body fat without affecting hippocampal neurogenesis and learning relative to an isocaloric glucose diet

Study: https://www.nature.com/articles/srep09589.pdf

Energy Macro Percent Usage
DietCarbsFatsProtein
Glucose62.5%18.3%19.2%
Fructose62.3%18.2%19.5%

Historical charts show the Fructose fed mice had more weight gain, and were less active than the Glucose fed mouse.

One Constant Macro*****Cumulative change of all macros is 0.6%
Constant Nutrients + Fiber*****Vitamins, Minerals, and Fiber was the same.
Number of animals per diet***12 animals per diet
Length of Study*77 days
Funding Bias*****Grant from Center for Nutrition, Learning and Memory
Total3.8Fair Quality Study

Study Review: High saturated fat and low carbohydrate diet decreases lifespan independent of body weight in mice

Study: http://europepmc.org/backend/ptpmcrender.fcgi?accid=PMC3922950&blobtype=pdf

Energy Macro Percent Usage
DietCarbsFatsProtein
Control Diet60%15%25%
High Fat Diet15%60%25%

Historical charts show the high fat fed mice had more weight gain and a shorter lifespan.

One Constant Macro*****Protein stays the same
Constant Nutrients + Fiber*****"Both diets were formulated in our laboratory and contained standard vitamins and minerals mixed with all essentials nutrients"
Number of animals per diet****12-13 animals per diet
Length of Study*****27 months
Funding Bias*****Grants by Brasil-Swiss Program and CNPq
Total4.8High Quality Study

Study Review: Severe NAFLD with hepatic necroinflammatory changes in mice fed trans fats and a high-fructose corn syrup equivalent

Study: http://www.physiology.org/doi/pdf/10.1152/ajpgi.90272.2008

Energy Macro Percent Usage
DietCarbsFatsProtein
Standard Chow (2018S)58%18%24%
Fast-food (TD.06303)37.1%45.3%17.6%


One Constant Macro**Protein changes by 6.4%
Constant Nutrients + Fiber*Completely different vitamins and minerals.
Number of animals per diet***8-10 animals per diet
Length of Study**16 weeks
Funding Bias*****Grants by Saint Louis University Liver Center.
Total2.6Poor Quality Study