When I make a lobby to practice against bots, I've noticed that hard bots are harder than unfair bots. Specifically, if I fill all 9 empty slots with hard bots, it seems like no matter how well I play, it's very difficult to win. Yet, when I choose to fill the slots with unfair bots, I win almost all of the time, regardless of how well I play.

Any thoughts on how to set up a match so how well I play is more likely to determine the outcome? Should I try 5v4, mixing hard and easy or some other set up?

By the way, I've tried medium bots and found that they're too easy. They seem particularly weak at harassing me in the laning phase so it's too easy to get last hits. However, medium bots on my team tend to randomly do stupid things individually and in groups so my results are very erratic.

Thank you and best regards.

  • that's weird. In my case, "unfair" bots are much stronger, because they're more likely stick together rather than the "hard" bots. Anyway I don't know how to make an even match with bots, but I suggest you to play it online with human :)
    – hendryanw
    Commented Sep 16, 2014 at 2:54
  • Actually, the same thing happens on CS:GO. The hardest bots are easier than the second hardest bots.
    – peper757
    Commented Sep 16, 2014 at 5:28
  • You will never be satisfied with the bots. They are either godlike and cheat on purpose, or they are terrible and feed in a single file line. I have played hundreds of bot matches (I need to pause often), and some things just do not work.
    – StingyJack
    Commented Sep 14, 2015 at 17:02

3 Answers 3


This is an opinion, but try to put a medium level bot in your team and lane with him. Mixing more easier bots in your team can lose one lane easily, leaving you unlikely to win except you are playing carry. As you say 4vs5 is also interesting. You will have more gold to invest since the gold income of the fifth player will be divided within your teammates.

I think the ultimate difficulty is playing with other humans. Some won't know how to lasthit/deny, teamwork, buy the right items, play roles... let alone understanding English, even if you play in the US servers.

  • I agree that playing humans is the true test of one's skills, and it's really where the fun is too. However, as a noob - with many heroes yet to play and learn - I'd really like to be able to use the bots in a way that helps me measure/monitor my improvement. Maybe it's not possible given the tremendous variation in Dota matches.
    – ScottO
    Commented Sep 17, 2014 at 13:30
  • you can play with players against bots. when i used to learn a new hero i usually did some coop vs. bot games to get a feeling for his skills and for a decent item build. if you're just looking to improve in something specific (which you should do if you want to improve in general) then you need to do specific things. for last hit training there is a tutorial. to train playing support, you need to play with and against other players. if you want to try out stuff, but don't want to wait for a game to try it, open a lobby, activate cheats and then just do it. Commented Sep 25, 2014 at 7:59

For good learning purposes, I always put hard bots in my team and let the enemy team have unfair bots.

If you put any lower than hard bot, (e.g.: medium bot or easy bot), they will just show poor performances and teamwork. Obviously you cannot learn much from these bots.

Another trick to increase the difficulty is by recommending weak heroes to your ally's bot. For example, recommend pudge for bot since bots are not really good with this hero. (note they will always pick your recommended heroes if it is available)

With that being said, if you still find this mode too easy to win, then I recommend you to match with real human.


It would probably be best to keep the same number of bots on both teams. Try having two unfair and two hard against 5 unfair. Also, keep in mind that bots can be somewhat controlled by pinging the map. You can also play a game with a friend against bots to help balance the teams. Playing against bots is not an ideal way to measure improvement and testing performance over several games would be better. You may want to consider ranked matches as they measure improvement over time.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .