Part of the issue is there’s a disconnect from what’s being shown and what’s already happened. So, XCOM, and I think XCOM2 (it’s been a while since I played both) create a table with “random” values on map load. This means, you can 100% save scum the shit out any encounter because cause and effect will always be the same, it’s not a live “dice roll”. Part of this sucks, because what happened is hidden from the player. Something like BG3, you can see “Oh, I swung, rolled a 3, and these modifiers, my total was 14 and they have an AC of 15”. Also, some games help by using a pseudo-random where the probability of something happening, actually increases over time. Example would be Dota2, where something like bash, shows a given percent, but it’s actually on a scale. Each attack changes the % chance the next bash may happen, eventually getting to a point it’s nearly a guarantee. This type of random is often used to make the game feel more fun for the player (to nudge the numbers one way or the other). However, with a pre-seeded table, this likely isn’t happening.

Then you add the visual component. Point blank range, it’ll say “99%” and you miss. Or the number will seem low, despite point blank range. And you have the visual of the %.

So you add those together, the game likely not helping the player and just using a pre-seeded table plus the visuals with the human notion of really only remembering the extremes and you get the overall feeling of “game not fair”. You made 10 shots in a row with only 30% chance, but you only remember the single 99% chance you missed