What you with criticals is utterly random.

Before, weapons gave up a bunch of damage in order to get each additional crit pip, and keen/improved critical where somewhat balanced based on being used on the better weapons.

Your changes:
Code:
Before   Before keen After         After keen
20x2 5%  19x2 10%    20/x3.5 3.5%  19/x3.5 10.5%
19x2 10% 17x2 20%    19x3 6%       17x3 20%
18x2 15% 15x2 30%    18x2 10%      15x2 21%
20x3 10% 19x3 20%    20x5 4%       19x5 12%
20x4 15% 19x4 30%    20x6 5%       19x6 15%
(The percent values above is a rough indicator of the boost to damage.)

Nearly randomly reorder what the values of the crit ranges are. If you are doing this, you might as well completely throw out the D&D 3.5e weapon tables, because any semblance of balance is gone.

And that 3.5 multiplier? That's just ugly.

Take a look at the 19x2 and 18x2 rows above. Under standard d20, 18x2 is better at generating critical damage than 19x2 -- quite rightly! Compare the scimitar to the longsword: they are similar weapons, except the scimitar loses a die size, and goes from 19x2 to 18x2. That boost to crits is supposed to be worth something.

Yet under your system? The difference between a keen 19x2 and a keen 18x2 is 1%!

In effect: you should first figure out what effects you want to change in your system. Decide "I want crits to be less important" or "I want the difference between crit multipliers to increase with non-keen weapons, and stay about the same when they are keen", and then tweak the mechanics.

It appears you are simply tweaking mechanics, and then accepting whatever random side effects fall out. That is a bad way to tweak a game system.

Find out what properties of the game system you want to change, and then figure out how to change the rules of the game system to get those property changes. Don't change the rules for the sake of changing the rules.

So, honestly, before: did you intend to decrease the impact of crits? Did you intend to make a long sword average boost to damage from crits be indistinguishable from the scimitar after you keen both? Did you intend to nerf the 20x3, 20x4 weapons significantly in average damage delt? Did you consider the keen scimitar to be too good?

All of these things you have done -- and it appears they are just happening as side effects of rule changes, as opposed to pre-meditated goals that you aimed to change rules in order to produce.

...

So what is the table at the end of my post? It is an attempt to produce a critical table that gives the same average boost to damage as the d20 critical tables. Ie: if your goal is to not change the relative quality of weapons, then that is the sort of table you want to use.

I'm advising you to take that approach -- compensate for changes done by default, and only after consideration change them.

...

Oh, as a rather neat side: if you also change D&D to a contest based game (ie, each side rolls 2d10 plus modifiers, instead of one side rolling and the other side providing the target), then this exactly compensates for the decrease in SD!

Ie,
1d20+ATK vs 10+DEF
is nearly identical to
2d10+ATK vs 2d10+DEF
in terms of the probability impact of having a +1 or -1 edge against the opponent. :)

The only significant differences are in the low-probability zone -- ie, out among the +/- 10 edge to one side or the other area, in which d20 often just gives up and says "you have a 5% chance to win".

Doing
1d20+ATK vs 1d20+DEF
has effect of making modifiers 1.4 times less important (or 70% as important), if you want to go the other way. This "flattens" the level curve somewhat, and might single-handedly reign in "save or die" spells.