what is the difference between mistakes and blunders?

a blunder loses immediately.   a mistake can be more subtle -- and should take more effort from your opponent to get an edge

@horsesgalore idk if that's the best explanation. A blunder can be subtle as well, but if opponent finds best move, u will lose immediately. A mistake won't lose immediately, but will put u at a big disadvantage.

Moves are marked as blunders when you give away a significant advantage.  That doesn't mean it loses immediately, you may still win in the end, you just lost the big advantage and clear win you could have had

In computer analysis, the difference between a "mistake" and a "blunder" is determined by size of the difference in computer valuation between your move and the "best" move

A mistakes is any kind of suboptimal move.

A blunder is a moves that costs you the game

I thought an inaccuracy was a suboptimal move.

A blunder only counts as such if your opponent notices it  I've got away with quite many in my time. That horrible sinking feeling you get when you realise you've just played a real stinker. Then you have to pray he/she doesn't see it...

Blunders do not necessarily give your opponent a potential advantage if it is a missed win however on analysis it is still marked as one.

It's all subjective. One annotator's (or programmer's) mistake is another annotator's inaccuracy. 

For the simple minded like me, a mistake is a move that causes me problems.  A blunder is a move that causes me major problems (usually significant loss of material).  Even this is subjective.  What is a problem versus a major problem?  What is a significant loss of material?

A blunder does not mean you lose immediately. It's a SIGNIFICANT mistake essentially. A VERY bad move. I can make a blunder and still win a game. It all depends what your competitor does and how you "regroup" and recover. In the computer evaluation scores you will also see blunders reflected as large values eg. +3 etc 

If you were at a critical juncture in the match where a win was in sight and you make a serious blunder it often says "from winning to losing". Again that doesn't mean you are 100% dead in the water...you could still recover...but it would be exceedingly hard work to recapture the massive advantage you gave away. 

It's mostly engine terminology nowadays anyway.  You get a whole bunch of armchair experts instantly saying a GM blundered because their little engine bar suddenly changed lots.

A blunder is relative to the level you are playing at though.  If you have two 1200s playing there is no expectation that they are going to be finding computer moves so labelling every other move a blunder is not really helpful to anything because there is an equally low chance their opponent will find the move that capitalises on it in many cases.

Blunders should be determined after the game as moves which had a significant effect on the game at that point.  People who want full engine eval to see the best moves they could have played in every position an get that anyway,

@Chess.com I think inaccuracy is less than 100 centipawns , mistake is between 100 and 200 centipawns , and blunder is 300+ centipawns  .  not sure what the official limits are but it's close

Or if it's the quick analysis at the end of live games they just make it up.  I went forward and back a couple of times to one of my games after finishing it and it counted different numbers of blunders and mistakes each time!

that's disconcerting .  I wonder if maximum analysis also wobbles ?

A blunder in game analysis by an engine means you lost a significant number of Pawn Equivalents - I think as natarka pointed out a couple comments before me.  You could still be winning, but if you were ahead 10.50 and your moved reduced you to 7.50, you'd still be winning but the engine considers it a blunder.

Don't get too caught up in what the chess engine says.  In one endgame, I moved my King into the Opposition position, guaranteeing my Pawn would Queen and I won several moves later.  The engine said it was a blunder!  Apparently, there was a faster way to win - but which would be tricky for a human to play mistake-free.

Personally, I prefer to save the pgn file of the game after chess.com analyzes it (using the share icon) and then load it into the freeware Lucas Chess and have Stockfish 10 at 20-ply analyze it, It takes about 70 seconds per half-move on my 3 GHz quad-core processor, Win 10 laptop. You can then look at each move and see other moves and how the compared.  You also get a little different set of indexes than the faster analysis at chess.com.  For example, I'm Black in the following game and won it in a tough struggle where a sac'd Pawn left me -1 in material but my opponent resigned.

In the tables below, note that chess.com says White made the best move 52.4% of the time, but Stockfish10 20-ply says White made the best move on only 3 of the first 19 moves then four of the last 5 (29.2% total).  chess.com say Black made 45.5% best moves and Stockfish 10 roughly agrees with 10 out of 24 (41.7%).

Notice that Chess.com's CAPS rated White 83.68 (about 1675 ELO) and Black (me) at 75.67 (about 1100 ELO) but Lucas Chess/Stockfish-10 20-ply rated White 2066 and Black 2797 and assessed that I dominated the game much more than White, which was the case.

Here's chess.com's analysis:

Here's Lucas Chess/Stockfish 10 20-ply analysis:

Here are the individual Stockfish 10 20-ply move ratings. Best moves are in Blue, ?? are blunders and ? is questionable.  On move 23, White grabbed a Poison Pawn:

WOAH!

@MickinMD - where is chess com's CAPS tool? I can't find it.

I just saw it. It's just some numbers, I guess.

A mistake is asking the forum a question.  A blunder is actually expecting relevant answers.  Hope that helps.