Friday, November 06, 2015

Elo elo elo, what's all this then?

There's been some talk among all the top memory people about rating the Extreme Memory Tournament competitors, using something along the lines of the Elo system in chess. Or for that matter in othello, which uses basically the same principles.

Until I looked it up last week, incidentally, I always thought it was the ELO system and the letters stood for something (although probably not Electric Light Orchestra), but it turns out it was named after a man called Elo. Anyway, the basic idea is that your rating goes up by a bit if you win a game and down by a bit if you lose. How big the bit is depends on the difference between your rating and your opponent's.

Now, the othello ratings are calculated by David Haigh, who's the only person in the universe who knows how it works. Well, him and all the people who've read his detailed explanation of how they're calculated and understood it, but I don't fit into that category. I've got a vague idea, but that's about as far as it goes. Not daunted by this, I've crunched some numbers and drawn up a rating list to see what it would look like.

People start out with 1400 points, for no good reason except that that's what the othello ratings start with, and increase/decrease by an amount that is 16 points if the memorisers have identical ratings, tending towards 0 or 32 depending on how big the gap between their ratings is. The change goes on a normal distribution (I don't entirely know what that is, I'm not a mathematician, but I put numbers into Excel's normal distribution formula until it comes out with the same kind of numbers as David quoted in his article). The othello ratings work differently for people who've played few or no matches before, but since everyone is starting from scratch here, and since I don't understand the calculation of provisional othello ratings (no, seriously, I'm not a mathematician, I'm an accountant and financial analyst, it's an entirely different discipline) I haven't done that.

Anyway, this arbitrary ranking makes a list that looks like this:

ALL Name Matches Rating
1 Simon Reinhard 61 1616
2 Johannes Mallow 73 1601
3 Boris Konrad 55 1537
4 Jonas von Essen 54 1509
5 Christian Schäfer 44 1503
6 Alex Mullen 36 1497
7 Katie Kermode 25 1476
8 Enhkjin Tumur 25 1470
9 Johannes Zhou 21 1409
10 Marlo Knight 15 1407
11 James Paterson 12 1399
12 Gunther Karsten 12 1395
13 Andi Bell 12 1385
14 Mark Anthony Castaneda 45 1381
15 Akjol Syeryekkhaan 19 1375
16 Marwin Wallonius 32 1374
17 Ben Pridmore 38 1371
18 Erwin Balines 12 1369
19 Yanjaa Altansuh 20 1366
20 Lance Tschirhart 15 1357
21 Ola Kåre Risa 37 1354
22 Enkhmunkh Erdenebatkhaan 15 1337
23 Tsogbadrakh Saikhanbayar 15 1320
24 Annalena Fischer 31 1314
25 Anne Reulke 15 1313
26 Norbert Reulke 15 1306
27 Bat-Erdene Tsogoo 12 1293
28 Johann Randall Abrina 27 1287
29 Tuuruul Myagmarsuren 15 1279

Obviously with there being so few matches for some people (like James, who won six out of twelve in 2014 but didn't qualify for the knockout stages), it's not super-scientific, but I think it still makes interesting reading. I would've hoped to be higher than 17th, since I basically made up the rules myself here, but that's how it worked out.

Incidentally, if you didn't know, you can see all the matches in the wonderfully-interactive XMT Live sites. Here's 2014 and here's 2015. If you want to see all Simon's matches, for example, type 'Simon' into the bar at the top. If you want to see all of Anne's images matches, type 'Anne images'. And so forth. It's great!

But the thing about the XMT is that there are five different kinds of match - six, if you count the surprise task. Some people are good at some disciplines but bad at others. Really, if you want to analyse who's good at the XMT, you might want a ranking system for each discipline. So even though there are so few matches as to make the whole thing pretty meaningless, that's basically what I've done here...


Cards!
CARDS Name Matches Rating
1 Simon Reinhard 10 1505
2 Ben Pridmore 8 1488
3 Mark Anthony Castaneda 10 1478
4 Johannes Mallow 15 1473
5 Alex Mullen 8 1465
6 Christian Schäfer 7 1418
7 Lance Tschirhart 3 1417
8 Enkhmunkh Erdenebatkhaan 3 1416
9 Gunther Karsten 3 1416
10 Jonas von Essen 11 1409
11 Yanjaa Altansuh 4 1402
12 Enhkjin Tumur 4 1401
13 Ola Kåre Risa 9 1388
14 Marlo Knight 3 1388
15 Anne Reulke 3 1387
16 Andi Bell 3 1387
17 Akjol Syeryekkhaan 3 1385
18 Bat-Erdene Tsogoo 3 1384
19 Erwin Balines 3 1384
20 Norbert Reulke 3 1382
21 Boris Konrad 10 1380
22 Johannes Zhou 4 1371
23 Marwin Wallonius 6 1368
24 Tsogbadrakh Saikhanbayar 3 1356
25 Tuuruul Myagmarsuren 3 1356
26 James Paterson 3 1355
27 Annalena Fischer 7 1352
28 Katie Kermode 4 1345
29 Johann Randall Abrina 6 1344

That's more like it! Second place in the card ranking! I have tended to win all my matches there, which is why it's always my first choice in the knockout events. Possibly even against Simon, though I'll have to think about my strategy there. The interesting thing is that Mark Anthony comes in third - I never really thought of him as being a cards specialist, but looking back at his results, he's very consistently good at it. And also improved quite a lot between 2014 and 2015, too. Be worried next year!


Images!
IMAGES Name Matches Rating
1 Johannes Mallow 8 1517
2 Christian Schäfer 5 1443
3 Boris Konrad 6 1435
4 Katie Kermode 6 1431
5 Simon Reinhard 5 1415
6 Enhkjin Tumur 5 1415
7 Mark Anthony Castaneda 3 1415
8 Alex Mullen 6 1407
9 Akjol Syeryekkhaan 4 1403
10 Marwin Wallonius 4 1401
11 Ben Pridmore 4 1400
12 Johannes Zhou 4 1399
13 Annalena Fischer 4 1397
14 Jonas von Essen 4 1397
15 Marlo Knight 3 1387
16 Norbert Reulke 3 1385
17 Yanjaa Altansuh 3 1385
18 Enkhmunkh Erdenebatkhaan 3 1384
19 Lance Tschirhart 3 1384
20 Tsogbadrakh Saikhanbayar 3 1384
21 Ola Kåre Risa 3 1355
22 Anne Reulke 3 1353
23 Johann Randall Abrina 3 1353
24 Tuuruul Myagmarsuren 3 1353

Did you know Hannes won eight out of eight images matches in 2015? See, the cool thing about these ratings is that they show that kind of trivia. Enkhjin was unquestionably the fastest when it came to images this year, but the XMT is about beating your opponent and making sure you get it right every time.


Names!
NAMES Name Matches Rating
1 Simon Reinhard 14 1542
2 Jonas von Essen 14 1512
3 Akjol Syeryekkhaan 3 1446
4 Christian Schäfer 9 1443
5 Ola Kåre Risa 9 1428
6 Marlo Knight 3 1419
7 Enhkjin Tumur 5 1417
8 Andi Bell 3 1416
9 Erwin Balines 3 1416
10 James Paterson 3 1415
11 Johannes Zhou 5 1411
12 Anne Reulke 3 1404
13 Alex Mullen 5 1404
14 Katie Kermode 4 1403
15 Johann Randall Abrina 6 1401
16 Tsogbadrakh Saikhanbayar 3 1398
17 Yanjaa Altansuh 4 1389
18 Tuuruul Myagmarsuren 3 1384
19 Enkhmunkh Erdenebatkhaan 3 1383
20 Marwin Wallonius 7 1377
21 Johannes Mallow 15 1373
22 Annalena Fischer 6 1373
23 Mark Anthony Castaneda 11 1365
24 Lance Tschirhart 3 1358
25 Boris Konrad 12 1357
26 Bat-Erdene Tsogoo 3 1353
27 Gunther Karsten 3 1353
28 Norbert Reulke 3 1352
29 Ben Pridmore 9 1307

Oh, come on! I'm not THAT bad at names, am I? Really? Sheesh. Let's move on.


Numbers!
NUMBERS Name Matches Rating
1 Christian Schäfer 10 1449
2 Johannes Mallow 12 1447
3 Enhkjin Tumur 5 1447
4 Katie Kermode 4 1433
5 Alex Mullen 6 1430
6 Andi Bell 3 1416
7 Gunther Karsten 3 1416
8 James Paterson 3 1415
9 Tuuruul Myagmarsuren 3 1411
10 Ola Kåre Risa 7 1409
11 Jonas von Essen 9 1409
12 Boris Konrad 12 1408
13 Johannes Zhou 4 1406
14 Johann Randall Abrina 6 1401
15 Annalena Fischer 6 1401
16 Ben Pridmore 8 1398
17 Yanjaa Altansuh 4 1397
18 Akjol Syeryekkhaan 4 1397
19 Marlo Knight 3 1387
20 Enkhmunkh Erdenebatkhaan 3 1387
21 Norbert Reulke 3 1385
22 Lance Tschirhart 3 1385
23 Simon Reinhard 13 1384
24 Tsogbadrakh Saikhanbayar 3 1384
25 Erwin Balines 3 1384
26 Bat-Erdene Tsogoo 3 1383
27 Marwin Wallonius 7 1368
28 Anne Reulke 3 1353
29 Mark Anthony Castaneda 9 1310

Hmm, here's a weak spot for the unstoppable Simon, it seems... although kind of a weak spot for me too, so taking advantage of it might be a problem. Numbers is the easiest discipline to make a mistake in and fail to get the perfect 80, so it probably favours people who are just the right amount of careful...


Words!
WORDS Name Matches Rating
1 Simon Reinhard 11 1545
2 Johannes Mallow 14 1534
3 Jonas von Essen 11 1520
4 Boris Konrad 9 1477
5 Christian Schäfer 10 1452
6 Katie Kermode 5 1420
7 Johannes Zhou 3 1416
8 James Paterson 3 1416
9 Gunther Karsten 3 1400
10 Alex Mullen 7 1394
11 Mark Anthony Castaneda 8 1391
12 Lance Tschirhart 3 1390
13 Ola Kåre Risa 7 1387
14 Marlo Knight 3 1387
15 Annalena Fischer 7 1385
16 Erwin Balines 3 1384
17 Norbert Reulke 3 1382
18 Tsogbadrakh Saikhanbayar 3 1381
19 Enhkjin Tumur 4 1377
20 Yanjaa Altansuh 4 1374
21 Akjol Syeryekkhaan 4 1370
22 Marwin Wallonius 7 1366
23 Ben Pridmore 7 1360
24 Anne Reulke 3 1358
25 Enkhmunkh Erdenebatkhaan 3 1356
26 Andi Bell 3 1355
27 Bat-Erdene Tsogoo 3 1355
28 Tuuruul Myagmarsuren 3 1353
29 Johann Randall Abrina 6 1315

Now, if I could find a way to get 47 every time in words, I could fight my way back up the rankings, maybe...


And finally, those Extreme Memory Tasks...
SURPRISE Name Matches Rating
1 Simon Reinhard 8 1480
2 Boris Konrad 6 1458
3 Alex Mullen 4 1430
4 Katie Kermode 2 1430
5 Marwin Wallonius 1 1417
6 Jonas von Essen 5 1414
7 Enhkjin Tumur 2 1403
8 Ben Pridmore 2 1399
9 Akjol Syeryekkhaan 1 1384
10 Annalena Fischer 1 1384
11 Christian Schäfer 3 1383
12 Yanjaa Altansuh 1 1383
13 Johannes Zhou 1 1382
14 Mark Anthony Castaneda 4 1371
15 Ola Kåre Risa 2 1369
16 Johannes Mallow 9 1312

Not really enough data here to go on, especially since the surprise tasks are different every time...


And so that's that. Comments, queries and suggestions are very welcome!

4 comments:

Simon Reinhard said...

Wow, great. We should do this with chess-like calculations, too. There the best would be much much farther apart, I assume.

Main problem surely is the low number of matches but it should still be possible to get this.

Clay Knlght said...

Very interesting to read.
How much more work would it be to make thew starting point 1,000 instead of 1,400?
Is it as simple as subtracting 400?

Zoomy said...

Simon - I'd have to find out how the chess calculations work and then create something similar, but I think it'd probably come out about the same.

Clay - yes, exactly that simple. If the starting point was 1000, everybody's current ranking would be exactly 400 points less. Othello started out using 1000, apparently, and then added 400 to everybody a few years later to stop a particularly unsuccessful player getting a negative score. :-)

Anonymous said...

Fist fighting to the death of one is the only true way of determining ranking.