If everyone just plays to win, this will work. For new players, it will get them to the right rating faster. A problem only comes up if you have players who try to "game" the system. One technique is to lose by large margins when you lose and win by small margins when you win. Only a few players will try to run this sort of scam. If you're worried about it or suspicious it is possible to look at the win/loss margins for each player.
One point to note is that you should probably do the arithmetic by percentages rather than ball count for the margin. If a player loses by 20 balls, it's different if he was going to 30 or 120.
Here's the way our spreadsheet works.
Let's say Bob played Mike and the score was
Bob 120 Mike 42
The spreadsheet notes that we played 162 points of straight pool. It then uses our rating difference to determine an expected score that adds to 162. Let's say the expected score based on our rating difference is
Bob 103 Mike 59
So Bob exceeded his expectation by 17 points, and Mike fell shy of his expectation by 17 points.
The formula that we use to update the ratings has some subtleties, but it is close to the following description.
Bob's rating increase is (4.2/MBOB) * 17
MBOB is the number of matches Bob has played with the caveat that we won't let it be smaller than 3 nor larger than 30.
So if Bob is a new player (new guy, first few games of the season), then his rating goes up by 24 points [(4.2/3)*17]. By the end of a 12-week season, Bob's rating would go up by 6 points instead of 24 with the same performance [(4.2/12)*17].
Mike's rating would go down by (4.2/MMIKE) * 17
We have data for 26 players who all played 14 matches within the group. We are able to compute the optimal ratings for the group after the fact. These optimal ratings don't depend on any updating at all. Then we are able to recreate the league week by week and see how fast various update procedures converge on the actual. Something close to the above procedure gets there pretty quickly. In other words halfway through the season it doesn't matter much whether we started with good guesses or started everybody the same.