Suhas Kashyap

I'm 1750 Elo on Age of Empires 2 now

February 16, 2026


That is roughly top 500-1000 in the world.
This post goes over mechanics of Elo-rating system rather than the story of how I got here.
If you're familiar with the Elo system in chess, this is very similar to that.

Quick definition:

  • "I'm" - Me, Suhas, who hit this rating about a year ago. My gaming PC has been dead since then.
  • "1750" - Big number. It is very likely that I will beat you in a 1v1 match.
  • "Elo" - Rating system used to rank / rate players in competitive games.
  • "on" - On the AOE2 ranked ladder - a competitive matchmaking system in which players are given a rating based on their 1v1 performance.
  • "Age of Empires 2" - AOE2 a real-time-strategy video game - where you play as different civilizations and defeat your opponent's army and base. Released in 1999!
  • "now" - My rating hasn't reduced but I might be a bit rusty.

Age of Empires 2 is my one true love. An embarassing amount time has been spent playing this, but zero regrets. It's the greatest of all time 🐐 because it's been here and thriving since 1999! The game has found new love with Microsoft's funding, and the competetive scene has become more and more active.

On AOE2 you start at 1000 Elo.
My 1750 rating puts me at 500-1000 out of ~50,000 players in the world. Quite the achievement I would say!

1v1 Random MapN = 50,000
0100200300400500600700800Number of Players1750 - I'm here!050010001500200025003000Elo Rating

Disclaimer: I'm not a mathematican, a lot of this info is lifted off of Wikipedia and Youtube.

What exactly is Elo?

"Elo" was invented by Arpad Elo, chess master and physics professor to calculate relative skill levels of chess players. A lot of zero-sum competetive games have adopted it now (physical and digital). Is not an acronym btw!

Simple idea:
Given two players, Elo system can predict the probability of a player winning over the other.
And, after a match, it can update the players' rating based on the outcome.

The number goes up when you win, and goes down when you lose. The amount depends on the difference between you and your opponent's ratings.

If you are 100 Elo over your opponent, you are expected to win ~64% of the time.
If you are 400 Elo over your opponent, you are expected to win ~91% of the time OR go 11 - 1 in a 12 game series.

Math

Two parts - the expected score and the rating update.

Expected score is the predicted probability of winning based on the rating difference.

EA=11+10RBāˆ’RA400E_A = \frac{1}{1 + 10^{\frac{R_B - R_A}{400}}}

RAR_A is your rating, RBR_B is your opponent's. This gives a number between 0 and 1: win probability.

After the match, your updated rating will be:

RA′=RA+KƗ(SAāˆ’EA)R'_A = R_A + K \times (S_A - E_A)

SAS_A is the result (1 = win, 0 = loss), K is the "K-factor" which controls how much ratings change per game. AOE2 uses K=32.

SAāˆ’EAS_A - E_A is where the magic happens. If you were expected to win (high EAE_A) and you do win, SAāˆ’EAS_A - E_A is small so you barely gain anything. But if you were expected to lose (low EAE_A) and you pull off the "upset", SAāˆ’EAS_A - E_A is large and you gain a lot of Elo.

Elo Calculator

Adjust ratings to see how much Elo changes:

Step 1: Calculate expected score
E=11+10āˆ’150400=0.703E = \frac{1}{1 + 10^{\frac{-150}{400}}} = 0.703
You're expected to win 70.3% of the time
Step 2: Update rating based on outcome
If you win (S = 1)
R′=1750+32Ɨ(1āˆ’0.703)=1759R' = 1750 + 32 \times (1 - 0.703) = 1759
New rating:1759
+9
If you lose (S = 0)
R′=1750+32Ɨ(0āˆ’0.703)=1727R' = 1750 + 32 \times (0 - 0.703) = 1727
New rating:1727
-23

Skill distribution bell curve

The idea with this normal distribution is that you might not always perform your exact rating; You might have off days or great days.

Drag sliders to see how expected score changes with the rating gap.
Overlap is where "upsets" happen:

100 point difference
13001400150016001700180019002000210022002300Rating
EA=11+10 100400E_A = \frac{1}{1 + 10^{\,\frac{100}{400}}}
EA=0.359935E_A = 0.359935
Player A wins ~36.0% of the time / (~1 win in every 3 games)

Self correction

The biggest draw of this system is that it is self-correcting. You could keep winning against the same oppoent (their rating unchanged) and your rating would keep inflating but the rate would get slower and slower - until you lose a match to get the inflated rating back down.

Winning again and again versus weaker opponents is heavily penalized, while going up against a stronger opponent is rewarded even if you lose.

Me at 1750, clapping a 1200, only gives me +1 Elo for the Win. It'd take me forever to climb the ladder! But if I were to lose, I'd drop down by 31 Elo. 31 games worth of hard work gone!
Likewise, if I improved and were to beat a 1900, I'd gain 23 Elo. But a loss here would only cost me 9 Elo.

Try simulating this by clicking on wins and losses against opponents of different ratings. You start at 1000 Elo:

Match Simulator

800900100011001200

Notice how a loss against a noob makes you lose all those puntos that you worked so hard for. Hurts.


That's about it.
Let me know if you think I could help explain this better or create better visualizers.
I'll write about how one could get to this level in a future post. Perhaps when I hit 2k!