The Gambling Effect

Discussion in 'Clicker Training' started by TeamGSP, Dec 15, 2013.

  1. TeamGSP

    TeamGSP Registered Users

    Joined:
    Dec 9, 2013
    Messages:
    528
    Ok I'm sure most people are aware of the gambling effect aspect of training where we gradually begin to reduce the frequency or level of reward given to a dog for performing a particular task. It feels to me to be a very human attribute to be attributing to a dog. And personally it raises a question of when to treat and when not to treat.

    If my dog performs a task perfectly why would I withold a treat ? Just to increase the dogs addiction to performing the action ? :-\

    So my question is this

    If a dog is performing exactly as you want him to perform would you treat everytime ?

    Is the gambling effect only to increase the dogs performance, sitting faster to a command, holding a retrieve until instructed to drop etc.
     
  2. JulieT

    JulieT Registered Users

    Joined:
    Jun 15, 2013
    Messages:
    20,186
    Re: The Gambling Effect

    [quote author=TeamGSP link=topic=3676.msg41390#msg41390 date=1387118325]
    It feels to me to be a very human attribute to be attributing to a dog. And personally it raises a question of when to treat and when not to treat.
    [/quote]

    I'm not sure - it always seems odd to say "in the wild" when talking about domestic dogs. I'm quite sure there are no, and never have been any, wild labradors. ;D ;D ;D Anyway...I think the point is that there is no 100% result in nature - so chasing the rabbit won't always result in catching the rabbit etc.

    A trainer I went to for a while convinced me to phase out treats for things Charlie did well over a month - 20% a week. I have to say, I think it did work better than continuous treats. It also made the treat more valuable when I handed one out so it made the treat more powerful - which I suppose is part of the "gambling effect". Plus, made it easier to keep Charlie's weight under control. ::)
     
  3. TeamGSP

    TeamGSP Registered Users

    Joined:
    Dec 9, 2013
    Messages:
    528
    Re: The Gambling Effect

    Thanks Julie so you would say to phase out treats such as kibble etc and reward less frequently but with a higher reward such as some nice chicken or beef.

    I always wondered if your dog thinks it done something 100% and didn't get a treat does the dog think it didn't get something right ?
     
  4. JulieT

    JulieT Registered Users

    Joined:
    Jun 15, 2013
    Messages:
    20,186
    Re: The Gambling Effect

    I haven't heard of less frequent but higher value treats as part of a treat phasing out programme. Mainly people talk about swapping higher value for lower, then phasing out (never completely though). I suppose your point makes good sense in terms of keeping an occasional "jackpot", which could strengthen the gambling effect. So phase out to kibble say 20% or 10% of the time, but give a juicy bit of chicken rarely. Would probably be effective.

    When I started phasing out rewards for a few things, Charlie looked a little expectant at the point where he'd normally get a treat but at first it was only 1 in 5 where he didn't get the treat, so he wasn't too disappointed I don't think. Plus, he is still young, so we always have more difficult things to move onto, and he gets treats again when it gets more difficult.
     
  5. charlie

    charlie Registered Users

    Joined:
    Sep 29, 2012
    Messages:
    12,217
    Location:
    Hampshire, UK
    Re: The Gambling Effect

    [quote author=JulieT link=topic=3676.msg41400#msg41400 date=1387121921]
    I haven't heard of less frequent but higher value treats as part of a treat phasing out programme. Mainly people talk about swapping higher value for lower, then phasing out (never completely though). I suppose your point makes good sense in terms of keeping an occasional "jackpot", which could strengthen the gambling effect. So phase out to kibble say 20% or 10% of the time, but give a juicy bit of chicken rarely. Would probably be effective.

    [/quote]

    This is exactly what I do, I always have a 'Jackpot' treat in my pocket, usually sardines or sausage for that excellent recall. Hattie & Charlie always think they are going to get a treat for everything, but they don't and so always encourages a quicker response next time. I do the same with the stop whistle. Really works :)
     
  6. TeamGSP

    TeamGSP Registered Users

    Joined:
    Dec 9, 2013
    Messages:
    528
    Re: The Gambling Effect

    [quote author=charlie link=topic=3676.msg41401#msg41401 date=1387122163]
    [quote author=JulieT link=topic=3676.msg41400#msg41400 date=1387121921]
    I haven't heard of less frequent but higher value treats as part of a treat phasing out programme. Mainly people talk about swapping higher value for lower, then phasing out (never completely though). I suppose your point makes good sense in terms of keeping an occasional "jackpot", which could strengthen the gambling effect. So phase out to kibble say 20% or 10% of the time, but give a juicy bit of chicken rarely. Would probably be effective.

    [/quote]

    This is exactly what I do, I always have a 'Jackpot' treat in my pocket, usually sardines or sausage for that excellent recall. Hattie & Charlie always think they are going to get a treat for everything, but they don't and so always encourages a quicker response next time. I do the same with the stop whistle. Really works :)
    [/quote]

    This sort of response from people who have used the method is what I want to hear. So start high with the reward value and then reduce the reward value and frequency but always keep a jackpot on board.
     
  7. charlie

    charlie Registered Users

    Joined:
    Sep 29, 2012
    Messages:
    12,217
    Location:
    Hampshire, UK
    Re: The Gambling Effect

    Yep :) That's what Pippa's Total Recall taught me.
     
  8. JulieT

    JulieT Registered Users

    Joined:
    Jun 15, 2013
    Messages:
    20,186
    Re: The Gambling Effect

    That is useful, Helen - I had only read of jackpots in terms of training the behaviour in the first place, not as means to maintain it once established and treat faded out, but it make sense.
     
  9. Stacia

    Stacia Registered Users

    Joined:
    May 25, 2011
    Messages:
    6,924
    Location:
    Malvern UK
    Re: The Gambling Effect

    You would still praise the dog if it did well, if you withheld the treat.
     
  10. TeamGSP

    TeamGSP Registered Users

    Joined:
    Dec 9, 2013
    Messages:
    528
    Re: The Gambling Effect

    [quote author=Stacia link=topic=3676.msg41475#msg41475 date=1387141984]
    You would still praise the dog if it did well, if you withheld the treat.
    [/quote]

    Would I still click ? If using C&T but the the only thing removed is the food reward.
     
  11. JulieT

    JulieT Registered Users

    Joined:
    Jun 15, 2013
    Messages:
    20,186
    Re: The Gambling Effect

    No, the click is always a marker that a treat is on the way.
     
  12. Jen

    Jen Registered Users

    Joined:
    Aug 30, 2013
    Messages:
    2,897
    Re: The Gambling Effect

    Hi Julie beat me to it

    No you wouldn't click if you remove the treat. If you click you have to treat. Otherwise the click loses its 'power'.

    You don't have to treat instantly after the click but a treat should always follow a click. That much I do know ;D

    Otherwise I can't add much to what's already been suggested I'm afraid.

    Jen :D
     
  13. TeamGSP

    TeamGSP Registered Users

    Joined:
    Dec 9, 2013
    Messages:
    528
    Re: The Gambling Effect

    [quote author=JulieT link=topic=3676.msg41487#msg41487 date=1387144636]
    No, the click is always a marker that a treat is on the way.
    [/quote]

    All good advice, thanks everyone. Going to be my first time clicker training.

    So it's almost a non event bar some fuss for the dog and on to the next exercise :)

    So when clicker training the clicker takes the place of a verbal marker such as good etc. So my dog goes off on a retrieve brings me back the dummy to hand, I click when the dummy hits my hand and reward without any further verbal praise but I can give him a good fuss

    If I introduce verbal praise on top of the click is that just confusion or would you all verbally praise after the click and reward ?

    I have pippas book by the way I'm just getting personal experiences from you all
     
  14. Jen

    Jen Registered Users

    Joined:
    Aug 30, 2013
    Messages:
    2,897
    Re: The Gambling Effect

    I can't help myself saying 'good dog' it just comes out automatically. It doesn't seem to confuse them.

    I've only been clicker training about 3 months or so. My dogs were 2. The difference in there training and behaviour since using the clicker is amazing.

    I know the theory of how it works but Im still surprised by the difference the click makes rather than just giving a treat. They definitely seem to work harder to get that click.

    Jen :D
     
  15. JulieT

    JulieT Registered Users

    Joined:
    Jun 15, 2013
    Messages:
    20,186
    Re: The Gambling Effect

    [quote author=TeamGSP link=topic=3676.msg41503#msg41503 date=1387180103]
    So my dog goes off on a retrieve brings me back the dummy to hand, I click when the dummy hits my hand and reward without any further verbal praise but I can give him a good fuss
    [/quote]

    I've been thinking this over lately, and when I get chance will try to look it up properly. I think this "isolated click" is almost exactly the same as a "good boy click". To my mind, it's not really harnessing the power of clicker training although no doubt there is no harm in it if the click is at the right point in time. I think the real power in clicker training is the fast repetition of an exercise or string of exercises. I wonder, once the behaviour is trained, you may as well use "good boy" for something well done. And if the behaviour isn't trained, but you want to train it with a clicker, you should be doing the fast repetition exercises. I may have got this completely muddled by the way! ;D ;D ;D
     
  16. TeamGSP

    TeamGSP Registered Users

    Joined:
    Dec 9, 2013
    Messages:
    528
    Re: The Gambling Effect

    [quote author=JulieT link=topic=3676.msg41548#msg41548 date=1387206698]
    [quote author=TeamGSP link=topic=3676.msg41503#msg41503 date=1387180103]
    So my dog goes off on a retrieve brings me back the dummy to hand, I click when the dummy hits my hand and reward without any further verbal praise but I can give him a good fuss
    [/quote]

    I've been thinking this over lately, and when I get chance will try to look it up properly. I think this "isolated click" is almost exactly the same as a "good boy click". To my mind, it's not really harnessing the power of clicker training although no doubt there is no harm in it if the click is at the right point in time. I think the real power in clicker training is the fast repetition of an exercise or string of exercises. I wonder, once the behaviour is trained, you may as well use "good boy" for something well done. And if the behaviour isn't trained, but you want to train it with a clicker, you should be doing the fast repetition exercises. I may have got this completely muddled by the way! ;D ;D ;D
    [/quote]

    I started reading total recall for the 2nd time last time lol

    Need to get this into my head in the next 5 weeks before the pup arrives.


    So when recall training and the dog out doing its thing do you click the instant it responds to the command or click when it's back at your feet, this is of course after initial shaping and the command being learned.

    Where's PIPPA ! Lol
     
  17. pippa@labforumHQ

    pippa@labforumHQ Administrator

    Joined:
    May 10, 2011
    Messages:
    5,466
    Re: The Gambling Effect

    The gambling effect is not specifically a human attribute at all, but simply an observation which recognises that unpredictable and intermittent rewards may have a more powerful 'reinforcing effect' on animal behaviour.


    I have heard it stated, that the value of intermittent rewards compared with a continuous schedule of reinforcement (or treating every time) has been called into question. But I have not yet seen any research to back this up (that is not to say that there isn't any). So in principle I always fade valuable rewards once a skill has been grasped, and before moving up a level.
    Not really, this type of training objective would normally require continuous reinforcement, because the dog is still in the learning process. The gambling effect is more useful when you are 'fixing' a learned behaviour to maintain it for the long term
     
  18. TeamGSP

    TeamGSP Registered Users

    Joined:
    Dec 9, 2013
    Messages:
    528
    Re: The Gambling Effect

    Thank you

    I have been hearing that continuous reinforcement is being promoted as a viable alternative to the gambling effect, that's one of the main reasons I asked the initial question to see if anyone had the evidence to dispute the gambling effect.
     
  19. JulieT

    JulieT Registered Users

    Joined:
    Jun 15, 2013
    Messages:
    20,186
    Re: The Gambling Effect

    [quote author=TeamGSP link=topic=3676.msg41566#msg41566 date=1387215489]
    [quote author=JulieT link=topic=3676.msg41548#msg41548 date=1387206698]
    [quote author=TeamGSP link=topic=3676.msg41503#msg41503 date=1387180103]
    So my dog goes off on a retrieve brings me back the dummy to hand, I click when the dummy hits my hand and reward without any further verbal praise but I can give him a good fuss
    [/quote]

    I've been thinking this over lately, and when I get chance will try to look it up properly. I think this "isolated click" is almost exactly the same as a "good boy click". To my mind, it's not really harnessing the power of clicker training although no doubt there is no harm in it if the click is at the right point in time. I think the real power in clicker training is the fast repetition of an exercise or string of exercises. I wonder, once the behaviour is trained, you may as well use "good boy" for something well done. And if the behaviour isn't trained, but you want to train it with a clicker, you should be doing the fast repetition exercises. I may have got this completely muddled by the way! ;D ;D ;D
    [/quote]

    I started reading total recall for the 2nd time last time lol

    Need to get this into my head in the next 5 weeks before the pup arrives.


    So when recall training and the dog out doing its thing do you click the instant it responds to the command or click when it's back at your feet, this is of course after initial shaping and the command being learned.

    Where's PIPPA ! Lol
    [/quote]

    I don't remember a lot about using a clicker in total recall - which bit do you mean, exactly?

    I wouldn't use a clicker particularly when recalling a dog, although I used to click for speed - so used the clicker to mark a precise point which I later rewarded.

    Here is a kikopup video that shows repetitive recall exercises using a clicker. I think the point here is the dog is being clicker trained in exercises, rather than having learned the behaviour and being rewarded with a "good boy C&T" for getting it right. Hope that helps. Probably not! ::)

    http://www.youtube.com/watch?v=PL9Rk-8KF9I
     
  20. TeamGSP

    TeamGSP Registered Users

    Joined:
    Dec 9, 2013
    Messages:
    528
    Re: The Gambling Effect

    Everything helps Julie ! Thanks
     

Share This Page