Is playing against bots an effective way to improve?

I am new to chess and have been working on getting better. I play puzzles daily and online games as-well. I also have been working my way through the chess.com bots. Are these bots good practice? Or is my time better spent playing online games, at the sake of my Elo? 

This is one of my better games I've played based on accuracy and analysis. Is this a good game or should I put an asterisk by it because of the bot-opponent?

In general it's better to play humans because they're bad at the same things you are. Not only does this make the errors more instructive, but for example a human will know how to fight back in a way that makes it hard on you. Humans also tend to play at the same strength in all areas all game long. Engines tend to make very stupid tactical or strategic mistakes and then play accurate tactics to make up for it.

But engines can be good in at least two cases:

1) if playing humans makes you nervous and so by playing against bots you're getting more practice than you normally would

2) if when you play humans you tend to play for tricks, hoping your opponent doesn't see your threat. If this type of player plays a bot that can beat them 70-80% of the time it's a good way to break that habit.

(playing a bot that can beat you 100% of the time is not very good since you lose the habit of checking if your opponent's move was a mistake, and overall develop a more passive play style)

No, weakened bots err, but in a non-human way.
A full strength bot is a good sparring partner, but you always lose.

I use the engine for middle game plans and preparation; basically see what kind of middle game I am getting?

Do I enjoy it or is it playable?

I would rather spend 5 minutes with the engine then trying to burn the clock in a live game to find out.

Your rating is not self worth points, it is simply a number that finds good opponents for you.  You won't get anything from having a high chess.com rating, or a low chess.com rating.  While you can use it to track progress, you can get a lucky streak of opponents that bump it all over the place.  I would consider it garbage points. 

To answer your question though, I've found both bots and humans mixed are the best.  

Bots cannot play like a bad human, because they are not humans.  They are programed to blunder, and they do very unhumanlike blunders, like random king moves. 

Humans blunder because it looks good to a human, but they missed something, like how a piece is hung or forget to check a long diagonal.  Often times you need to practice finding human blunders, so a game like that is good. 

However, bots do two things better than human players at low levels.  You can't play "hope" chess with mates.  Even Jimmy, who will hang his queen on the regular, will move out of a mate in two blunder.  To mate a bot, your mate game has to be on point, which is a good exercise. 

And two, low level bots will go for draws if they can't win.  Low level human players will often play known draws to the end, or even blunder a known draw rather than take the draw.  Bots see they will lose and get that forced 3 repetition. 

Yes, bots help.  The beginner ones like nelson are a good teacher, once you can beat them regularly, you will have improved.   Try the ones rated above 1000, as bot ratings are inflated and the ones below 1000 are doing little more than making random legal moves. 

I've been exclusively playing against Nelson for the past few months, along with puzzles and free lessons once a week, and I feel like I'm improving, albeit slowly. Bot training does work, no matter what GMs tell you.

This is a common myth, but I think it is wrong.

Human players make every kind of mistake imaginable, and so do weak bots.

@9
I believe it is possible to tell if an opponent is human or not from how he errs.

You can learn from any bot.  I think beginners get a LOT more from them, those early lessons of not hanging pieces or getting swindled give you a lot for a little time spent.   

Playing challenging chess will improve your game. Period. I play bots when I don't feel like having time pressure and when I don't want to imagine some human at the other end of the line gloating over my allowing a knight fork. But I also play humans a lot. Mix it up, but, above all, play chess that makes you really think. I've been working my way up through the bots (with no help, hints or takebacks), and I just beat the Isabelle bot the other night on the first try. Now it's on to the next bot, and the next, until I meet my Waterloo. I never move up, though, until I've beaten the one I'm on. Now, today, I'm headed to an OTB tournament--7 rounds at 15/3 per round. Should be exciting. And exhausting.

You will get better by playing bots of increasing difficulty; however, it is more effective to play human opponents.  The bots are basically full-strength engines with a random error rate.  So, you will see a bot play perfectly tactically for dozens of moves and then hang an obvious mate-in-1 or hang the queen, etc.  Humans do not play perfectly and then make a crazy blunder (we play poorly and then make a crazy blunder!).

But like I said, you will still improve by learning to recognize the errors when playing bots.

The lower rated Bots are horrible and the game you posted in the first post is a perfect example. That was a terrible game by the computer. There's virtually nothing instructional in there. The bot either hung or didn't capture 15 points of material by move nine.

I can only agree that playing humans a lot, AND analyzing the games for mistakes you may have made is the best way to improve. (I often think I played a great game, and during analysis I was blundering along as usual, but was just lucky my opponent didn't see them ).

I only play bots when I really don't want to be stressed in any way. I don't know about you, but playing a human always excites me, which also makes me more attentive with my moves. Against engines, I tend to be really lazy and not even caring when I blunder. Also, at maximum rating, playing engines at length is frustrating, because they'll always squeeze you to death in the most sadistical way. And I found the engine's play at lower levels really ridiculous -- human blunders have still an inner logic that you can learn something from, but engine blunders are just more or less random moves. (I know some people disagree, but that's my impression.)

For opening preparation, playing a top engine is fine to see where you opening may have hidden long-term weaknesses, but even then, I do prefer to use a game database statistics. Be careful not to use only "Master Databases", because you will virtually always play against patzers like you and me. Also, you have to put in LOTS of work to understand the WHYs of top-level opening theory moves, and I have always fared much better if I play "my" openings with moves I understand, leading to positions that I like. I do look at engine evaluations, but again, I found that opening database statistics are much more valuable in finding devious moves to confuse and unsettle, disturb and/or annoy my opponents.

Which brings me to my main point: Human chess is also a very psychological game. If you are feeling confident with a move or position and your opponent doesn't, you have already nearly won. Making uncomfortable moves, even if they are not the top engine moves, will make your opponent nervous or impatient, and sooner or later he will make mistakes.

But in the end, I'm just here to play chess. Of course I'm happy if my rating gets better, but I don't mind losing one game after the other, wasting 100 rating points because I have a blundery day, as long as I enjoy the games. And I think that ultimately, the thing that makes you better in chess is practice and experience, training your brain to slowly see hidden patterns, harmonies and imbalances in a position. Reading a good textbook also helps. Just take care you use something that's about your level, not too high-level and not too low-level either.

Play unrated! A million times better than bots

I feel like using bots exclusively is not good. I can beat 1000s bots, but against people, I barely scratched 500s and I average 450 elo. They make nonsensical mistakes that I have never seen a human make.

This is normal.

Bot ratings are not really ratings (they don't change with wins and losses), those are just numbers that some developer figured out, and almost always they are way off.

Also, I have never seen a nonsensical mistake that humans don't make. Humans do all sorts of mistakes.

Humans do all sorts of mistakes, but what you do not often see is someone playing 15 perfect moves, blundering their queen to a 1-move tactic, then playing another 30 perfect moves.  That is how the bots play because they are 3600+ engines with forced errors