Start at 10K. Draws are unrated. Ratings change via formula 250 - number of loser’s moves. Combined change is always zero. No floor or ceiling. You’re very welcome 
Yes, please. Let’s give really bad players one more reason to play on until mate!
Alex Relyea
…And let me introduce the highest rated player in the country … a kindergartener that knows scholar’s mate. 
Zero sum ratings systems only work if a pool remains stable, but as people who gained points leave they take those points with them. And that’s just one of the many mathematical flaws in this concept.
Still, the US Chess ratings system is not the be all and end all of ratings systems, despite what those on the relevant US Chess committee may believe.
If anyone believed that, we could dissolve the committee. Instead I believe it is constantly trying to improve the system.
Alex Relyea
I’ve never anyone on the ratings committee make that claim. There will be a ratings workshop in Orlando. You can come and ask them.
This, of course, is (I hope!!!) an intentionally unserious suggestion. Someone else (who shall remain nameless) has a penchant for “serious” suggestions which are either unworkable, or create collateral damage, or (generally) both.
Churchill said that democracy is the worst form of governance, except for all the others.
I’ve not seen a chess ratings system I feel is superior to the Elo-based systems. The differences between US Chess, FIDE and Glicko2 are minor, mathematically, and each of those systems have some advantages over the other two, and some disadvantages.
Elo-based ratings are used in a number of other contexts, not all of them sports.
Other than that, is there anything wrong with them? 
Two great advantages to the FIDE system are that a measure of experience and competency are required to have your play effect anyone else’s rating and that how the outcome of your game will effect your rating is definitively known (or knowable).
The former is politically impossible for U.S. Chess because players and parents would simply not stand for tournaments after which a large number of the participants remain unrated.
The latter is not politically impossible, but very difficult and would require a lot of cultural changes to have all games rated monthly instead of tournament by tournament. For example, rapidly changing players might find their rating change too rapidly. If an 1800 player who has a 1200 rating plays five tournaments in a month, his rating might go up 200 points the first, 120 the second, 80 the third, 40 the fourth, and 20 the fifth. If all events were rated at the same time, his rating would go up 200 points for each, so he’d be 2200 instead of 1660. Note very large ballpark estimates, but it should be obvious that a 1640 player will have far less of a rating increase than a 1200 with an identical performance. Also, there are organizers who like to use ratings at a time (maybe many different times) other than the monthly cut-off for pairing and prize purposes. This would be much more difficult, if not impossible.
Alex Relyea
Now that’s funny…
+1
“I understand there would be…no math” is probably not the ideal attitude for a discussion of ratings. In fact, you need to play a truly staggering number of games in a month for FIDE’s batching system to cause an overshoot. In the case of a 1200 playing 1800 chess, that’s around 36 to get the rating up to just 1800 (assuming the player has the highest K=40 value). And remember that the downside is that each one of those opponents in that month will have their rating computed as having played a 1200. Under US Chess ratings, a player starting at 1200 with 1800 p.r.'s in five round tournaments would go 1385, 1529, 1626, 1683, 1714, 1734, 1749, so by roughly halfway through this busy month, the opponents will be rated as playing someone who is fairly accurately rated.
The fact that FIDE’s rating system can take a 1200 with 1800 performances and (possibly) produce a rating in excess of 1800 is a mathematical flaw, but one that FIDE is clearly willing to accept (a) they don’t have bonus points and (b) the activity level required for the ratings to go “unstable” is hard to achieve—note that for the more common K=20, it would be twice as high. (US Chess avoids the potential instability of the fixed K Elo adjustment by adjusting K for the number of games rated.)
Yes, there are a handful of players for whom the FIDE ratings changes are greater—they use K=40 for players 17 and under who are rated under 2300, so a young player rated between about 1900 and 2300 would see quicker changes under FIDE (but would get to 1900 quicker, probably much, much quicker, under US Chess ratings).
The major differences between FIDE and US Chess ratings are not mathematical, they’re procedural, for example what it takes for someone to have a ratable result and get a published FIDE rating.
Also, FIDE uses the same pre-event rating for an entire month’s worth of results, US Chess does not. That has advantages, in that the impact on someone’s rating from an event is far easier to predict, because everyone’s pre-event rating is known.
One disadvantage is that rapidly improving players don’t see correspondingly rapid ratings changes. It also makes it difficult to use ‘most recent event’ ratings, but since FIDE doesn’t permit that, it isn’t an issue for them.
And for the most part, FIDE doesn’t have rapidly improving players, by the time someone starts playing in FIDE rated events, they’ve usually already made that leap of intuition or whatever it is that separates higher rated players from the rest of us.
Tom, if you look at the leader boards (uschess.org/datapage/leaderboards.php), you’ll see that it isn’t at all unusual for US Chess players to have 50 or more regular rated games in a month. That would be very difficult to do in FIDE rated events, because FIDE rules limit how many FIDE ratable games one can play in a day.