Rating Inflation: 100 Elo points starting in 1985

 I think the jump is actually true due to advancing knowledge of theory, communication of ideas and computers. Even aggressive players got better at calculating because they knew when to do so. Positional play and older masters like Morphy and Fischer would get smacked around by Carlsen, Karpov or Kasparov. But I think it would be cool if we could measure each players potential through some method. Let's say give 1 elo point to a player for the number of years between the present and the player's year of peak rating all the way until 2900. We should stop at 2900 because I don't think a human could be better than that.

I believe players nowadays are better thanks to the knowledge they get from internet and chess engines , but the rule  of counting a victory against a player lower than you with 600 , 800, or 1000.... Elo points equal to beating a player lower than you only with 400 points definitely must cause some inflation in the ratings, this is a big mistake the FIDE has made to encourage the best players to participate in low level tournaments, while they could simply encourage them by money without damaging the Elo rating credibility.

The rating system only compares players relative to each other. If we really think computers, databases, internet, etc. have improved the play of top players RELATIVE to lower rated players then that would explain some of it, but this seems unlikely as lower rated players have access to these things as well.

Honestly it's probably even more apparent for the lower rated players as they don't have to wait until chess club once a week to play real games etc. The top players still had intense training pre computer era whereas others had whatever random books they could get their hands on. Obviously computers/databases/internet have had a huge increase on chess training but it's done just as much if not more for the average amateur. 

ThaGoodBoy is probably right in that Fide's rule for 400 points caused bit of it. It should also be pointed out that this rule was introduced in 1980, so just a few years before we saw some big jumps. Pre 1980s the only fide rated players were very high rated as pointed out previously. It was just a different system, only top events got rated and the large majority of serious events a club player played in were unrated. This is why Kasparov started out with a rating of close to 2600, not that it was his first tournament but that it was his first one to be rated. Remember, ratings were done by hand in those days so it was much harder to have a large rating system than now. 

Another example is that for a while in the late 80s to 90s FIDE had the Karpov rule which stated that if you won a tournament you couldn't lose rating points. This of course inflated the system as well. 

Another example is rating floors which is common issue with the USCF rating system for example. 

Basically anything the messes with the "a point gained by one player is a point lost for another" can result in rating inflation or deflation over time. In the case of the Karpov rule (for example) points could be gained by lower rated players who beat or drew the winner but the winner didn't lose any points to account for that, therefore rating inflation. 

So the reason for all the rating jumps etc was because pre 1980s about 95% of chess that would be rated today was unrated and there have been many changes to the system which inevitably throw things off, it just so happens it threw it upwards. 

A way to refute the argument claiming it's just a an increase in accuracy would be that some players who played in the 60s-90s had higher ratings when they were old than when they were in their prime. Also, Kasparov's ratings in the late 90s/2000s would make him the clear #2 in the world right now, and the clear world #1 in mid 2017. So if you believe that there has been an improvement in chess accuracy over the last 20 years (and there probably has) this would make no sense that he would actually be similar rated to Carlsen. Therefore the ratings can't show the improvement in the accuracy of the games. 

I should add that the 400 points rule doesn't actually add points to the system the way the Karpov rule does but that it tends to help the higher rated player disproportionally, so could result in rating points being pushed/skewed upwards to some small degree. 

The way that players initial ratings are calculated can also have a serious effect on the entire system (especially over years and years). I'm not sure about the fide system but in USCF if you have a perfect score your performance rating is assumed to be 400 points above the highest rated player, (and for your first event your rating will be your performance rating) so a player can crush three 1100s and then beat a 1600 and be rated 2000 when they were really maybe 1750-1800. Though this rating is provisional and will likely change it still might take a few events especially if you continue playing in these low rated events, and your opponents can still get "credit" for beating a 2000 player. 

There is nothing complex about the math here. Let's analogize to a poker game where instead of an ELO rating we measure skill by chips. The game is organized by 10 players who each put 100 chips in the game. The average number of chips per player will always be 100, no matter how skilled or unskilled the players are. Some players will have more and some less, but the average will always be the same, and "chip inflation" is impossible. The same is true if one player leaves the game and another comes in and takes over the outgoing player's chips, even if the new player is a rank amateur and taking over for the most skilled player.

Let's consider a couple of possible scenarios: First, a new player wants to join. If he brings 100 chips to the game, the average stays the same. But if he brings 1200 chips to the game, there are now 2200 chips and 11 players, so the average is now 200 chips per player. The skill of the incoming player is irrelevant. The inverse is true if the new player only brings 50 chips to the game, in which case there are fewer than 100 chips per player and therefore the average number of chips has decreased.

From this we can deduce a rule: if the incoming player brings more or less than the average number of chips per player, there will either be chip inflation or deflation as the case may be.

The same holds true for chess. If a new player is rated at exactly the average rating of existing players, it will have no effect on overall ratings. Skill is irrelevant.

Now, let's consider a different scenario in our poker analogy. The players agree that it's unfair for a player to lose chips if the winner has a royal flush and the player has something under a full house. So a new rule for dividing chips is made: the winner gets to keep the pot, but the player with a qualifying hand does not lose any chips. Now, this is physically impossible unless additional chips are put into the system. Assuming chips are added, chip inflation results.

This illustrates how a rule change can affect the total number of ELO points in the system.

Obviously, the ELO rules are much more complex than a simple poker game, but the principles are the same. If points are added to the system that increases the average number of points per player, there will be rating inflation.

It has nothing to do with the skill of the players. It's just math.

A couple more points that are relevant to the conversation. When we think of rating inflation, we are looking at the ratings of top grandmasters, which are indeed higher than their predecessors. Unless we compare the average number of ELO points per player now with the average at some point in the past, we cannot determine if there is rating inflation or if there is just a bigger gap between the grandmasters and the rest of us.

None of this is remotely relevant to a conversation about the strengths of today's top grandmasters and those of past decades. Furthermore, I think it is pointless to talk about advances in chess theory, the role of computers or anything else that is external to the player himself. If we're going to compare Carlson with Fischer, aren't we really talking about a head to head match? In other words, if we want to compare Fischer to Carlson, we have to posit that Fischer and Carlson are the same generation, in which case Fischer would have available all the aids that current players have, or, alternatively, we could posit Carlson to be in Fischer's generation and have none of the resources he grew up with. To me this is obvious but I haven't seen it in any of the discussions I have some across, which are admittedly not very many.

The most accurate way to compare players of different generations is really to measure their performance against their peers. How much better was Fischer than his peers compared to how much better Carlson is to his. I don't pretend to know how to do this, although the gap between their ratings and the ratings of their peers would surely be relevant.

I think something is wrong with the spaces between your lines.