I dont think the calculalation uses intergers since it has decimals.
Exemple from my forcast  namr irrelevant and 0.5 is result of game

2286 (77)      2421 (64)      0.0434      -0.0572      0.5

There is a slight deflationary pressure simply from playing a game. This pressure is less as rating increases and RD decreases. I will document my calculations because they can change the calculator at any time.

---

Example calculation.
For Rating=2200 and RD=100, plug in a win against an opponent -200, a draw against an opponent =0, and a loss against an opponent +200. Use the same RD for player and all three opponents.

    My Rating                  2200      100 
    Opponent One Rating        2000      100  Win 
    Opponent Two Rating        2200      100  Draw 
    Opponent Three Rating      2400      100  Loss 
    Calculate                  2199.029   96.973 
    Change                       -0.971   -3.027 

---

Varying Rating. 
For constant RD=100, plug in a win against an opponent -200, a draw against an opponent =0, and a loss against an opponent +200. Use the same RD for player and all three opponents.

    My Rating      R-Calc      R-Change      RD-Calc    RD-Change
    2600  100      2599.162     -0.838        97.655     -2.345
    2550  100      2549.153     -0.847        97.575     -2.425
    2500  100      2499.125     -0.875        97.493     -2.507
    2450  100      2449.107     -0.893        97.409     -2.591
    2400  100      2399.090     -0.910        97.324     -2.676
    2350  100      2349.074     -0.926        97.238     -2.762
    2300  100      2299.058     -0.943        97.151     -2.849
    2250  100      2249.043     -0.957        97.063     -2.937
    2200  100      2199.029     -0.971        96.973     -3.027
    2150  100      2149.016     -0.984        96.883     -3.117
    2000  100      1998.982     -1.018        96.607     -3.393
    1800  100      1798.954     -1.046        96.231     -3.769
    1400  100      1398.963     -1.037        95.479     -4.521

---

Varying RD. 
For constant Rating=2200, plug in a win against an opponent -200, a draw against an opponent =0, and a loss against an opponent +200. Use the same RD for player and all three opponents.

    My Rating      R-Calc      R-Change      RD-Calc    RD-Change
    2200   17      2199.982     -0.018        16.984     -0.016
    2200   25      2199.961     -0.039        24.950     -0.050
    2200   50      2199.826     -0.174        49.606     -0.394
    2200   75      2199.541     -0.459        73.692     -1.308
    2200  100      2199.029     -0.971        96.973     -3.027
    2200  125      2198.200     -1.800       119.263     -5.737
    2200  150      2196.971     -3.029       140.431     -9.569
    2200  200      2193.052     -6.948       179.150    -20.850
    2200  250      2187.008    -12.992       213.097    -36.903

---

I already found one reason not to trust the Ratings Calculator. According to 2.1.b on page 151, the value RD used in the formulas is pegged at 30. This means entering any value less than sqrt(30^2 - 25^2) == 16.583 in the calculator should give the same rating change, but it doesn't.

The Ratings Calculator only accepts integer inputs. The formulas only make sense with floating point maths, but the document doesn't state whether Ratings are saved as integers or floats. It makes a slight difference because if saved as integers then small rating losses will be discarded and the deflationary pressure on the pool will be much less.

Dink Heckler wrote on 02/28/24 at 09:18:28:

One would hope so. Of course it is theoretically possible to construct an asymmetrical formula where the rating changes don't converge, but who would do that?

I took a look at the ICCF rules, at least the ones I could see without logging in to ICCF. 
https://webfiles.iccf.com/rules/2023/2023%20Updated%20ICCF%20Rules%20-%204-1.pdf

The formula(s) there can't possibly be the one(s) in use. On page 152:

• Probability of a win is a function of the player's own rating.
• Probability of a draw is a function of the average of the two players' ratings.
• Probability of a loss is a function of the opponent's rating.

None of the probability functions depend on the difference in the two players' ratings. Professor Elo is turning over in his grave.
Edited:

>Most high rated starter long before todays computers standard thus they where very good Chess players since more human work was needed.

But this is historic, it doesn't reflect the current results gap, which is marginal. The only logical outcome is players converge to similar ratings - it's not the function of the rating system to reflect how good people were decades ago. But for perfectly understandable human reasons, historically strong players resist this. Is this a fair summary?

Rating depends on opponents one mets rating and  it depends on start Elo = Fide Elo. High rated usually only plays in closed events where they met similar rating. And all other mets each other. After rules changes draws vs higher rated gives less than before.200 elo higher draws gives 1 or 2 elo.  I have met GMs and SIMs etc who has played a lot in open events and lost lots of rating this way. An other rating factor is when players started to play. Most high rated starter long before todays computers standard thus they where very good Chess players since more human work was needed.

I mean, if everyone is drawing all their games, shouldn't everybody ultimately converge on the same rating? 

Before the new system was introduced the forecast with old system said a slight minus. But when new system calculated i lost even more points so it depends a lot on opponents rating if losing more or losing less than in old system. I think the rating system has both inflation and deflation at the same time. Its not hard to draw 2600 but hard to win over 2100 if 2100 is using latest tools and plays mainstream openings or playing black vs drawish lines..

trw wrote on 02/26/24 at 03:48:23:

Dunno. Should this be the case?

The new ICCF rating system seems difficult to understand to put it mildly. The recommended calculator only allows three games, but maybe it can be of some help.
https://ratingscalculator.azurewebsites.net/default.aspx

Have a nice day.