I just played a game and it was an amazing game in my eyes (particularly the end).(M.B.Raho vs. Otonomos),
However, I lost, and I’m fine with that.
My issue is that my opponent and I were very close in ELO, I started with +4.5 Komi giving M.B.Raho a handicap. On top of that I lost, due to timer which AI agrees that I was hugely in lead and was even about to win (“Variation: White wins” in the chat.), and then I lost 11 ELO, which took ~3-4 wins to earn.
I just want to confirm this is all as it should be, as it feels a little harsh.
I played a previous game against a weaker opponent and they were given a handicap, then I gained 1 ELO. I really don’t understand the system.
I have a little goal in my head of hitting 700 before I really get into 13x13, but the system is making it very hard for me to do this. I’m sure ill get to 700 at some point, but dang the system is making it harder for me.
Does anyone better understand the ELO system and can help me to understand it?
To nitpick, OGS uses Glicko-2. Very similar to Elo, but not the same.
To uber-nitpick, it’s Elo, not ELO
Glicko/Elo/OGS rating system does not care about your predicted winrate during the game. It only cares who won or lost.
Were the 3-4 previous matches also evenly matched? If, for example, you were playing weaker players, it is expected that you’d need more wins to make the rank rise.
My recommendation is not to try to predict your ratings points with such precision. Here is a very high level way to look at it:
if there is a “surprising” outcome: your rank changes a lot
if there is an “unsurprising” outcome: your rank changes a little
What is a surprising outcome? It could be a 12k beating a 3d. Or a 5k beating a 5k with a 5 stone handicap.
If you are 15k and you beat a 25k on an even game, your rank will not change much because this was a predictable result. The system cannot really tell you are better than 15k. Playing evenly ranked games gives a much higher signal to the system.
Finally, you might be able to understand the system better by playing with @shinuito’s rating calculator: