Maayan

[Maayan @MaayanOrner]

@ganja_dr420 @mikemo @yutohorigome 95 was too low, it was the most shocking competition trick I have seen in my life. It should have been 97 because the gap is huge for a Nollie 180. It's just more nuanced to see how impossible this trick is.

[Maayan @MaayanOrner]

@JimSatala @jakejakeny @cooperunion And please publish the list, so everyone knows to never hire them.

[Maayan @MaayanOrner]

@mwfowlie @CixLiv @amasad Or using graphs. With their data, it's next to trivial in many cases.

[Maayan @MaayanOrner]

@alizac41007815 @urbanicGray @yairlapid שאין לאופוזיציה מספיק כוח בוועדה למינוי שופטים ושהרוב ל"עוקף בג"ץ" קטן מדי, מה שפוגע באיזון רשויות באופן קיצוני בעיניי (בכל מקרה הממשלה שלנו היא גם הכנסת).

[Maayan @MaayanOrner]

@urbanicGray @yairlapid דבר בשם עצמך. אזרחים רציונליים רוצים פתרון פרגמטי, לא לנצח בהורדת ידיים (שאגב לא תנצח בה כי אין לך רוב). גילוי נאות - אני חושב שהרפורמה מטומטמת וצ��יך לנסח אותה מחדש, ז"א לשנות את רוב הפרטים.

[Maayan @MaayanOrner]

@syaccoby @amit_segal אנקדוטלית לפי ההתרשמות שלי נראה שהתשובה היא כן.

[Maayan @MaayanOrner]

@YagaoDirac @ayonzontop @ylecun Now try to modify some layers to do something very specific and the fwd pass and let me know how it goes ;) That's the reason I like Pytorch, everything is more transparent and makes sense.

[Maayan @MaayanOrner]

@AmchBur @kumiiisrael אני לא יודע אם הבעיה היא מקצועיות של הטלפנים, משהו אחר, או שזה מכוון, אבל הסקרים האלה תמיד מפספסים. אני מקווה בשביל מפלגות השמאל (או כל מפלגה, לא לוקח פה צד פוליטי) שיש להם סקרים פנימיים יותר מדויקים, כי זאת אינדיקציה מטעה. מישומה נראה שהתיקון הוא פשוט להוסיף קצת לימין באופן קבוע

[Maayan @MaayanOrner]

@AmchBur @kumiiisrael נראה שהם לא יודעים לדגום כמו שצריך, תמיד כביכול איכשהו הם לא לוקחים בחשבון שהמון ימניים לא ירצו לענות לסקר של ערוץ 13, זה קרה בבחירות בארה"ב כבר לפני 5000 שנה, מעניין שעוד לא למדנו.

[Maayan @MaayanOrner]

@loedru12 @IglesiasYosha I liked the transparency, the implementation (and open source contribution) & presentation but I don't think his statistical analysis was of high quality, he tried to torture the data to show some predefined conclusions. In my opinion, of course.

[Maayan @MaayanOrner]

@loedru12 @IglesiasYosha refer to that reddit post https://t.co/LzN5p5b01p

[Maayan @MaayanOrner]

@loedru12 @IglesiasYosha Untrue, there are no anomalies - I validated it myself using the same method Yosha did and https://t.co/6P1ERNQGec came to exactly the same conclusion in the report.
I don't care if Hans cheated or not, but fake science is dangerous.

[Maayan @MaayanOrner]

@xenobyte_ @IglesiasYosha I will elaborate, Magnus has only 2 100% games (the sample size for Magnus is smaller), so if he would have only 1 more 100% game by accident (which is not unlikely to happen) it would change the number completely. Makes sense?

I guess no one caught him yet if he cheats OTB.

[Maayan @MaayanOrner]

@xenobyte_ @IglesiasYosha 0.82 against GMs, which means it's actually lower against high-level opponents.
And yes, for this sample size it's not significant - you can almost consider it noise.

[Maayan @MaayanOrner]

@IglesiasYosha On average, Magnus has a much higher "Engine Correlation".

The data and analysis are available here https://t.co/mC3nrniMBL

Also, it's a very simple and brief analysis, but it's enough to show the quality of your analysis.

[Maayan @MaayanOrner]

@IglesiasYosha I compared how many "100% games" Hans has in comparison to Magnus and Keymer and found that the numbers are very similar, Hans has ~1.28 per 50 games, Magnus has ~1.05, and Keymer ~0.92.

However, against GMs, Hans has only ~0.82 games per 50 games.

[Maayan @MaayanOrner]

@IglesiasYosha If you are interested, here are CSVs for 3 players and a starter notebook (kind of) https://t.co/mC3nrn1bdb

[Maayan @MaayanOrner]

@IglesiasYosha Could you publish a CSV with an analysis for other players? It's much more efficient than calling the critiques of your methods conspiracy theorists. If you claim something is wrong, you have to show what's ok, and you need to show how probable it is that it's "wrong" by luck.

[Maayan @MaayanOrner]

@NikolaosNtirlis @IglesiasYosha Also, I would bet my money on Ken Regan's algorithm which was probably designed carefully, rather than that pseudo-scientistic blitz scrumble of data.

[Maayan @MaayanOrner]

@NikolaosNtirlis @IglesiasYosha I agree, it's a terrible case of Dunning–Kruger effect, unfortunately.
As a data scientist, I am well aware I don't know much about chess, but the statistical analysis chess people perform makes me cringe :/