Replay search tools are not powerful enough

I want to watch high level games so I can see what higher rated players are doing to improve my own gameplay.

I am finding it quite difficult to search for high ranking games, because the current only way to try achieve this is the rating filter, but it works based off if any player in the game is within the requested rating.

So, if I were to set it to 2000, very often I would just get games where, for example in a 4v4, there are 1 or 2 about 2000 rated players, and the rest are just 1500 rated players. That isn't a very high level game, since most of the players are just 1500s. How 2000s play vs 2000s is potentially a lot different to how 2000s play against 1500s. I want to be able to search games where all the players are, say, over 1800 rating.

Being able to search rating by, for example, the mean rating of the game, or the median, or by the lowest rated player in the game would be incredibly useful for achieving this.

I am hoping the developers can add a drop down next to the ratings range filter to add these conditions, instead of just having the default search by any player in the game being within the requested rating.

I recommend the YouTube channel of JaggedAppliance

Some of his videos are replay reviews (where he shows a replay and talks about what to do or what not to do). You can search for his videos that have "replay review" in their title.

Other videos are POV of him playing ranked games, so you can watch what he's doing. This can be even better than watching a replay yourself, because you don't just see what moves he makes, but you can see what he's choosing to look at. Part of getting better at the game is just getting a sense of what you should be looking at.

Also, I agree with you that it would be nice to be able to filter out games with lower-rated players.

One way you can easily do that is to search only for 1v1 games. It's rarer for someone high-rated to be in a 1v1 with someone a lot lower than them. Even if you want to learn about how to play team games, you can learn a lot from watching 1v1s (and my advice is, if you want to get good at FAF, first work on getting good at 1v1, here's a shameless plug for my guide )

bump

I looked into it yesterday. It's neither simple nor straightforward with the way our api works internally and the different ratings that could affect a single game.

Is there no way simple way to alter a function to filter greater than min rating in game?

I've found the search tools to work pretty badly in general. Searching for a replay that includes more than one playername, for example, hasn't worked for me in ages. Search has also been very very slow for as long as I can remember.

SQL database are not build for flexible queries on 1:n data (1 game n players).

It's also not designed to run fuzzy text search (a map or mod name where i don't know exactly where whitespace and apostrophes are).

But: most software can't. Look at this forum: it's running on a document oriented db but it has the same search limitations.

Most of the smaller applications in the world behave like that.

Search for people that only play 2k+ avg, thats the easiest solution i found

And you can search by game name, look for games with things in the title like "2k+" "2000+" "1.9k" etc.

Is the replay database schema published somewhere? The search function is pretty limited, but I can write my own SQL search filter. I have no idea what the available fields are. I'd need the table layouts.

Nvm, I found it!
https://faforever.github.io/db/relationships.html
Hopefully, this is up to date.
Not sure I can do a table join in the client. I need to figure out how to run an arbitrary query so I can play around with it.

For the OP's problem, searching for high-rated players is about the best you can really do.
There probably are not many 4v4 games where everyone is 2000+. There just aren't many players at that level.

@PViddy The problem can never be solved with a SQL join.

What happens in the API is this: perform the join, then filter.

So first you join a game and get a result like:

And now you filter rating>=2000 on all records of that table. So the result would be this:

It's easy to find a game where on participant matches a criterium.
But there is no row-based filter in SQL that could require all players to be >=2k.
For that you need either SQL subselects (which is not supported by our API logic for technical reasons) or we calculate somehting like min/max/mean rating during fetching the game in the API.

The latter is generally possible, but not so simple with our API logic, because a game participant has potentially multiple leaderboards, so it need to return these KPIs per leaderboard (even though in practice each game right now has only one leaderboard).