I am interested in the "elo economy" and slowly I formulate useful questions that help me to gain insights over it. See more here
One of those question is "is there really an inflation that will let the scores to explode?" , although the best score so far , 2285
, broke the previous record only after months. I was thinking that 2300 would have happened sooner - In February - but instead it is quite difficult to get there. So if there is an inflation it is not so big.
Therefore I started to see the elo as a currency exchanged with dynamic exchange rates.
So I thought if there is inflation, or better "the rich ones get richer", it is true that since the score (or elo) is not created out of thin air in an exchange
then the group of people with higher scores are harvesting the group of people with lower scores to get "richer", or higher.
So, focusing on GM, I collected a quite large sample of stats..
89396 Games completed between the dates: 2017-05-02 07:05:25 and 2017-06-17 09:08:56 , Paris time.
The I asked the database. How much is the difference of points exchanged between people with 1400 or higher score and people with 1300 or lower score?
An example to clarify what I mean.
PlayerA 1380 vs playerB 1420 : playerB wins gaining 8 points
playerA 1380 vs playerC 1410 : playerA wins gaining 12 points
playerD 1350 vs playerE 1400: playerE wins gaining 8 points.
The net trade, or difference, is "the group with 1400 or more points" gained 4 points (8+8 - 12 ).
Now it is true that a player can sometimes appear in the group of "lower than 1400" and sometimes in the group "higher than 1400" but it does not matter much because the flow of points will show nevertheless a trend. The same can be said for players that go inactive therefore holding their points outside the flow because no other player can gain or lose points against inactive players.
So in the sample checked there were 7906 games between player under 1400 and players with 1400+ points and the trade of points is in favor of those with 1400+ points. This group gained 4742 net points.
Then I asked the same for the 1500+ group (therefore players up to 1499 points counted as "poor" in this case). Still the net trade is positive. The group 1500+ extracted from the group 1499- , 794 points in 13464 games. One can see that while the flow is positive, it shrank a lot. It is not that a lot of points flow from 1499- to players above their score.
Then I asked, how is the flow between players 1600+ and players 1599- ? Still positive (so, points "flow" upwards). 917 points gained in 15491 games.
Then the surprise started. I know that the elo is awesome and it is not forgiving. If one has a score way higher than another player, he needs to win really a lot and to keep the score, not even increasing it, one has to lose a little. But I did not expect what I found.
The players of the group 1700+ actually lose
points to the players below them. There is no harvesting! They have lost 757 points in 15920 games.
The negative trend continues. (so, no harvesting)
1800+ vs 1799- : 1082 points lost in 14259 games
1900+ vs 1899- : 1388 points lost in 11534 games
2000+ vs 1999- : 1524 points lost in 7202 games
2100+ vs 2099-: 1192 points lost in 2703 games
2200+ vs 2199-: 180 points lost in 373 games (indeed at this time there is no 2200+ in the leaderboard, not even inactive)
So what one can derive from this and the rest of the observations in many other threads? Surely that if there is a player with a very high score compared to all the others, that player has outstanding tactics compared to the active
playerbase (not compared to the inactive or the future playerbase) but it does not mean that when there are more players with "high" scores (like many 2000+) the quality is lower.
This also means that to keep high scores is very difficult, since there is a negative trade of points from the high score regions (the sweet spot is around 1700 points. "Easy" to take juicy points from the top and from the bottom), and I would say that to keep an high score one has to "harvest" other high score players rather than low score players, otherwise, one can see, the trade will end up being negative and the score level will be quickly lost.
So if before I thought the process was like:
1400 gets farmed by 1500, 1500 gets farmed by 1600, 1600 gets farmed by 1700, 1700 gets famed by 1800, 1800 gets farmed by 1900, 1900 gets farmed by 2000, 2000 gets farmed by 2100.
Now it is more likely that 1800, 1900, 2000+ gets farmed by 1700, 1600 and 1500 while high scores are reached because 1800, 1900, 2000 exchange points between each other.
Anyway, more info on this in some time.