Results from my simulations (your expectations of games are wrong)

Yesterday, I wrote a program that simulates League game behavior. Basically, it measures how a game progresses on a more theoretical level than simply gold. Instead of gold, I use "Decision," which can be thought of as a coin flip with probability based on skill difference between teams and the resulting value of the coin flip dependent on the time in game. The details and assumptions can be cumbersome, so I'll link to the thread with all that nitty gritty here.

But onto my first thoughts.

More than likely, your expectations of how a game progresses are just wrong. I simulated 57 40-minute games and then categorized them into 5 types: Landslide, Comeback, Swing Game, Rally, and Normal and the technical definitions can be found in the linked thread above. While 57 is not that decent of a sample size, the spread of the proportions are large enough to make a few sweeping observations:

1. Early game matters. Out of 57 games, 41 were determined by the early game.
2. Most of the above are rally games (25/41); Rally games are considerably more likely than Landslides
3. You are much more likely to get a landslide than you are for a comeback, but it is much more likely that the enemy team rallies than it is for you to win unimpeded
4. Swing games almost rarely happen. If your definition of a good game is for both team to go tit for tat or to go back and forth, you will be sorely disappointed.
5. The previous 4 categories are extremely good at catching games; there were only 2 "normal" games (Read: "other").

My conclusions: You should not go into a game expecting teams to be trading objectives, kills, cs, or even trades themselves. That rarely happens. You may find that really exciting because it pulls on your heartstrings, but it's not expected by a long shot. What is much more likely is that one team is going to go on a streak early on. They'll get a triple kill at 10 minutes and probably take dragon. Some people will say that's gg, especially since they will likely do it again. But that's not will most likely happen. The simulations support Game of Throws. It's not just a meme; it's actually true. Note, that does NOT mean the enemy will lose most likely. What will probably happen is you will start going on a bit of a streak, and if it comes late enough, rally hard enough that you just barely lose or you actually win. The latter is significantly less likely, but still very possible.

This is what you should expect: you honestly should expect one team to snowball in some way and then stay snowballed. It's most likely going to happen. What is also very likely is that they mess up or you play well and suddenly the game is starting to shift in your favor. This is where your skill comes in. The people who can close out a game and take advantage of the swingyness of the game are going to be the winners a lot of the time.

Learning how to close games is the deciding factor in skill level and victories, it seems. I've always believed that comebacks don't fall out of the sky. It's true. The fact that there are so many more Rally game in this simulation than comebacks means that if both teams are playing equally as well the whole game, the expected case is that you will rally back and just barely lose.

That makes for some exciting stuff to watch and even play, but if you are ultimately wanting to git gud, then you need to: work on your early game, and learn how to close games effectively.

And finally, if you surrender, take a really hard look at the game, because it will more than likely start to turn to your favor just by playing equally as well even in the late game.

All questions, comments, and critiques are appreciated! Even the ones that tell me to go kill myself :D.