The Art of Prediction

The best trainers make it look easy. They appear to know what the dog is going to do before it does it and, when something changes, they seem to know immediately what they are going to do next.

The honest answer is that quite often we do know, or at least we have a pretty good idea. There is no mystical Zen ability involved, it comes from years of watching dogs, training dogs, making mistakes and seeing the same or similar situations play out over and over again. We are making educated guesses based on what we have seen before.

We are predicting.

Prediction itself is an art, but it is grounded in a very simple question:

What is the dog likely to do next?

I was thinking about this recently and it took me back to my military days. There were commanders who, when a situation changed, appeared to know exactly what to do almost immediately, like they had read the script. There were also those who needed a little time to work out what had happened, consider the options and then come up with a solution.

I don't necessarily think the first group were thinking faster. The good ones were often thinking ahead. They had already considered what might happen and what they would do if it did. They were constantly running possible scenarios in their heads, so when one of those situations actually occurred they weren't starting the decision-making process from scratch.

I think we do exactly the same thing with dogs.

The problem is that the dog doesn't stop while we think. If something happens and I then have to decide what I've just seen, work out what it means and decide what I'm going to do about it, the dog has already moved on. The learning hasn't stopped while I've been thinking and neither has whatever is happening around us.

Recently, while talking about this with some visitors from the Swiss Army, Stefan reminded me of the OODA Loop. It was something I hadn't thought about for years, but as soon as he said it the connection was obvious.

OODA is Observe, Orient, Decide, Act.

You observe what is happening, orient yourself to the situation, make a decision and act on it. The result of that action changes the situation and you go through the process again.

It makes complete sense and clearly has parallels with training and handling dogs, but I kept coming back to what happens before Observe.

If I wait until something happens before I start the process, I am already reacting to an event. What I actually want to be doing is considering what might happen before it does.

If I can predict a number of likely outcomes, I can also consider what I am going to do if one of them occurs.

That is where I have started using a much simpler version for myself:

POR - Predict → Observe → React.

Predict what is likely to happen.

Observe what actually happens.

React to what you see.

Prediction in practice. The toy has already left my hand before the predicted sit, allowing the sit action to be reinforced just after it happens.

Your reaction changes the situation, so you start again.

Prediction doesn't mean deciding what the dog is going to do and then trying to make what you observe fit your prediction. That would defeat the point of observing the dog in front of you. We are simply trying to be prepared for what might happen.

If I predict correctly, I already have an idea of what I am going to do. If the dog does something different, I react to what I actually see and that experience becomes another possibility I can consider in the future.

This isn't specific to detection work and it isn't just about formal training exercises either. I think it applies to pretty much everything we do with dogs. Whether we are teaching a new behaviour, fixing a problem, handling a working dog, walking a puppy or competing in a trial, there is always value in asking what is likely to happen next.

It has also made me think differently about timing.

Timing is regularly cited as one of the keys to good dog training and clearly it is important, but I wonder if what we describe as excellent timing is sometimes actually the result of prediction.

If I wait for the dog to do something and then start working out what I am going to do, there will inevitably be a delay. If I have already considered that the dog might do it and I have an idea of what my response will be, I can react almost immediately.

To somebody watching, that looks like good timing. It might even look instinctive.

In reality, I haven't necessarily reacted any faster. I just started thinking earlier.

I think that is a large part of what experience gives us. Every dog we train, every dog we watch, every mistake we make and every time a dog does something we didn't expect adds another possible outcome to the library.

Over time we get better at recognising what might happen next, and because we have considered it before it happens, we get better at knowing what we are going to do when it does.

Perhaps that is why the best trainers make it look easy. It isn't that they always know exactly what the dog is going to do.

They are just very rarely starting from scratch.

Tom Moir

Detection Coach