First off, this is not me complaining about running badly or anything, Im just curious about how to understand how best to interpret the numbers, and if they have much meaning at all.

Numbers over a very small sample ~70 games, 2,700 hands @ 2.50 180s on stars.

net won -102,000 chips, cEV diff 98,442

now what I gather from that is:

1. During all-ins, I have won 98,442 chips less than my expected value
2. My expected value over this sample is ~ -3600 chips, meaning that on average I haven't got my chips in good (I realize it's a small sample and there are other factors to consider - see 4.).
3. For any player, regardless of skill, cEV diff will trend towards zero... or at least, any player is equally likely to have a + or - cEV diff over a sample. This is essentially a measure of how your luck is going during all-ins (obviously there are other ways you could be running good/bad that don't involve all-ins)

What I'm reeeally not sure about:

4. Should a winning player expect to have a positive expected value over a large sample? (that is net won + cEV diff). On first glance you would assume that a winning player would be better at getting their chips in good and thus have a positive net EV, and I'm sure this is true in calling situations. However, there are also raising situations were we might expect to be -EV if called, yet the raise is still profitable because of the dead money if everyone folds. Examples of this might be a standard bvb raise, or an UTG raise with only 3-4 BB left etc.

Hope this didn't all sound to scattered. Help much appreciated and please correct me on any misunderstandings I appear to have on these numbers.

Thanks!

I think your understanding seems pretty good but this stat can be obsessed over and may be dangerous to follow.
When trying to narrow/nail your roi you can nudge the actual value a bit either +ve or -ve by interpreting using the cEV diff. To some small extent it can help to see if you have actually lost some of your edge early indicating a good time for some research and work rather than waiting for your actual roi to catch up.

I am also unsure of (4) I think it depends on many factors.

You can get some use out of any stat but often you have to filter it to make any real sense.
I would expect a good player to have a larger positive 'expected' when calling rather than pushing and for pushing maybe it should be perhaps slightly neg (depends on some factors though). When calling I would expect a good player to typically consider their edge and most often call slightly tighter than a shove so usually be ahead.
(Any uncalled push doesn't get counted in this stat so although neg you easily make up for this from the fold equity in pushing.)

In the FT stages it would depend on the size of the player stack - if you hit an FT with a the top stack you can dramatically increase your aggro level compared to being mid-stacked and hence get caught with neg 'expected'. and if mid-stacked icm dictates that you should be very positive on expected.

Overall, there is some uses for this but much more in filtered situations - eg, if you want to find out how well you play mid pairs with a 30 to 40bb stack then using the 'expected' values gained/lost is generally better but has to be interpreted well.

ok, sounds like we are pretty much on the same page. The more I think about it, it seems to me that a overly-tight break-even type player probably has highest all-in EV. This type of player would be waiting for premium hands and so would be a heavy favourite in many of their all-ins.

Would help if someone with a decent ROI over large sample stepped up and provided some of these stats. I realize that it's not critical, I'm just curious more than anything.