The OODA Loop and the Half-Beat | Hacker News Reader

The OODA Loop and the Half-Beat

thestrategybridge.org

109 pointssmacktoward31 comments

As a photographer I have observed that people tend to blink or move or change position, or gaze direction loosely in sync, meaning they don't all blink at exactly the same time but around the same time (when not posing).

If you observe and have somehow an off-tempo you can take pictures of groups of 3-5 people having a conversation where every face is expressive and all eyes are opened, which is otherwise difficult.

But, I have this one friend I can almost never photograph (again, when not posing) because it seems we are on the same "beat". He blinks or moves at exactly the same time when I press the shutter. It's really strange because it's the only person I know with whom this happens.

> But, I have this one friend I can almost never photograph (again, when not posing) because it seems we are on the same "beat". He blinks or moves at exactly the same time when I press the shutter. It's really strange because it's the only person I know with whom this happens.

Wait, did your uncle inherit you a black leather belt with silver buckle?

What does that mean?

I think the grandparent poster thinks he knows the root poster personally and trying to confirm via anecdote.

Or they are referencing some meme?

This phanomin has been seen in studies that looks at how people watch films

https://www.newscientist.com/article/dn17526-synchronised-bl...

I have to leave my two cents on this topic.

I spent a good amount of time writing situated agents in persistent worlds (aka bots for things like everquest). My experience was that the tighter the feedback loop the less the bots would get stuck in weird places. So when I first was exposed to the concept of OODA, it was very clear to me that the concept mapped very neatly to the work I'd done. All the bots I've written got better essentially the more frequently I applied the OODA loop.

You'll also find if you read the literature on situated agents, that robots got much better much faster once scientists started building them to react to the world, as opposed to the early days where it was assumed they'd need to build these massive models of the world for the robots who would then need to spend massive amounts of CPU time trying to decide what to do next.

What I think the article may be missing, when it talks about half-beats, is that ideally you really want to iterate over your OODA loop as fast as possible, so that you can make more decisions than your opponent. The first step is often to define your Action in the smallest timestep that you can. So if your opponent thinks one step is their smallest action. You should really be trying to define the smallest action as a half step. Now you can potentially make 2 decisions in the same amount of time. Then do this for each stage of the OODA loop. Find a way to Observe faster. Find a way to Orient faster. Find a way to Decide faster (even deciding to defer deciding until the next OODA loop).

At the end of the day, your sample rate and your real time deadlines dictate a lot about how an agent can act. And don't get me wrong, a good algorithm will always trump a high sample rate, but don't underestimate just how important a high sample rate is when talking about an OODA loop.

The point with this article is that it misses a much deeper application of OODA. And even that application is a fairly simple application. This article is really talking about one example manipulating the OODA loop of your adversary.

Having said that, AlphaGo and the deep learning starcraft bots look like they are starting to approach that kind of exploitation of the adversay's OODA.

Sure, except that the article was talking about changing your OODA loop to somehow disorient an opponent. And while that might be useful the same can be achieved by having so many more OODA loops than your opponent that you can simply choose to drop packets in order to take advantage of some timing benefit. Generally, when you're doing sampling data, it's much better to sample at the highest frequency possible, and downsample as needed. Even with dumb linear interpolation you'll still generally have higher resolution with less uncertainty than if you try to reduce your sampling to match some kind of signal.

I do like you bringing up starcraft though. Clicks per second can largely be mapped to the frequency of the OODA loop. When I was following the broodwarAI bots a couple of years ago, usually the better bots tended to have a much higher meaningful clicks per second than worse bots. This of course, isn't necessarily causative, since the better algorithm may spend more time on analysis so that it clicks less, but each click is more valuable.

But I think you are right, in that the bigger idea OODA was that the reason the pilot who came up with the idea originally was using it to describe why his largely inferior plane was able out dogfight enemy planes. That the plane he was in had more responsive controls, which allowed him to react faster to his opponent, and also to keep forcing his opponent to get stuck reacting to his actions. His basic tenet was that because his plane was more responsive, he was able to keep forcing his opponent to have to go back the start of the opponent's OODA loop every time he changed course, thus wasting precious time observing and reorienting.

You're not "changing" your own OODA loop. I think you are misunderstanding OODA. Your understanding of it is precisely what that article was saying is ... oversimplistic at best. One of the ways to beat someone with a faster OODA loop is to change conditions in a way where small divergances with what is observed and what is actual will start to stack on top of each other. Once the opponent starts panicking and flailing, you can lead him around pretty easily.

Col. Boyd wasn't just talking about fighter pilots. Much of what he was talking about have a lot of relevance to fighting in general. I find a lot of skilled martial artists breaking someone's tempo. Musashi and Sun Tzu have talked about it in each in their own way.

Put it this way. I have a friend who regularly beats people on first person shooters. He does not have fast reflexes. He just knows how people think and react and can manipulate them.

But I don't really care. It isn't as if I were going to go out fighting someone seriously anytime soon. And if I were, I wouldn't want to teach my adversary how to use OODA properly.

This. I've applied Musashi's principles in competitive re-enactment combat successfully. Faster is generally better, and there's a point where a raw speed difference is enough to win any fight. But manipulating the tempo of the fight so you can hit them on the off-beat is definitely a thing, and that might involve slowing down a bit to manipulate that tempo. Which is exactly what TFA was saying.

I think the point of the article is that even if you can't make your loop any tighter, there can be benefits to timing your actions so it disrupts your opponent's loop.

That's the fighting example. Fights generally develop a rhythm, where people trade punches. You can get a big advantage by striking when it's not "your turn", even if this means waiting.

Boyd describes the OODA loop as an analogue process, not digital. A person or organization continually orients, observes, decides, and acts. This is different than Deming's PDCA cycle which is a serial process. Speeding up the OODA loop requires the 'right' culture for the objectives, openness to new information, distributed decision making, common understanding of mission and priorities (Boyd calls this schwerpunkt),... In military terms, the goal is to disorient the enemy and thus disrupt their OODA loop while being able to function effectively in a confusing environment.

From experience, this is critical in a military setting. I find this theory difficult to employ in the technology/startup arena in most cases. That said, the elements which make for an effective and efficient OODA loop are invaluable when building an effective organization. In my experience, many companies miss opportunities because insightful information (observe) often comes into the organization from unexpected sources and does not get properly taken into account (orient); employees have not uniformly internalized the organizations priorities (decision); and decision making cannot be effectively distributed (action).

Having played a lot of strategy games in real time against human opponents, I am not sold on the "faster iteration" tactic re: the OODA loop, nor it's timing-based cousin presented here.

Or at least, I'm skeptical that that's the best way to frame a lot of problems.

A better way, IMO is to frame it in terms of mental models.

Some mental models are more efficient to think in, just as some programming languages are more efficient to write in. Many domains are amenable to study beforehand, allowing you to figure out the right model to think about the domain, effectively pre-compiling the Orient and Decide steps.

This can be a big deal for two reasons. The more boring one is that pre-doing anything will probably give speed gains. The much more interesting one is that you can get qualitatively different Orient/Decide steps. A buttload of contemplation, condensed into a few maxims, is usually worth a lot more than thinking slightly (or even two or ten times) faster in a critical situation.

You are right. The book Science, Strategy and War basically promotes the idea that the Orient stage of the OODA Loop is where mental models come into play which makes sense since how else do you Orient without having an idea of what you are Orienting against. i.e, if you are in a dogfight with a Russian aircraft what are the mental models that you have that what mental models do they lack.

Hmm, I'll have to find that and give it a read.

I don’t think anyone’s arguing for speed over domain expertise. But all else being equal, iterating more quickly is an obvious advantage. You get more chances to correct mistakes and adjust to new information. Your decision may be to continue with the same strategy, but having the capacity to make that decision is important in itself.

Take fencing, for example. At low levels, people often try to make up for poor technique with speed and strength. Like you point out, this is a mistake. But at the highest levels, where everyone’s technique is basically flawless, the winner is the fencer whose thinking is a step or two ahead.

That said I’m really talking about situations where you have an active opponent. With more static problems yeah, being correct is more important than staying on your toes.

Worth noting also that speed in this case means quicker feedback, which means learning and integrating more domain expertise during the same unit time.

OODA was developed for fluid environments, where strategy and tactics must constantly adapt to an uncertain world whose understanding is constantly being updated.

Video games don't work well with this model because in most video games a) information is perfect and b) results are deterministic. You don't have to update your understanding of the world constantly because your knowledge of the world is more or less complete.

In John Boyd's world, you couldn't "pre-compile" any decisions because he was working in the real world, not a game.

I think this is much more of a spectrum than you're implying. And perhaps more than I've implied above.

The reality is that the vast majority of real-life situations are hybrids[0]---some things are constant, others are new. The math we do changes, but math doesn't change. Boyd's OODA period may have been shorter than that of enemy combatants, but he still had to go to training. What is training if not pre-compiled thought? And importantly: in fluid environments where novel things are likely to arise, having a solid grasp on things that are "settled" makes it easier to integrate new phenomena.

Precompiled thoughts are part of the OODA loop.

> Video games don't work well with this model because in most video games a) information is perfect and b) results are deterministic.

Neither of these are true in the strategy games the parent comment refers to. They mention playing "in real time", and I interpret that to mean real-time strategy games. In those games, you generally have:

1) Fog of war. You can only see the play area near where you have units. Enemy movements outside of that region are hidden from you. So you are constantly discovering new information as your unit moves around, and having to deal with where you thought the enemy was being wrong.

2) Randomized damage. Most attacks don't do a fixed amount of damage and instead do a certain amount of "rolling the dice". You know on average how much damage X will do against Y, but not the results of any single hit until after it has happened.

RAND corp has a really interesting video on their website about how hyper-sonic intercontinental ballistic missiles brings the length of the OODA loop down to about 5 or so minutes which is a very short amount of time to make a decision to mobilize an insanely large amount of resources.

https://www.rand.org/multimedia.html#search/hypersonic/multi...

It was about quickness not just raw speed. It's both about the rapidity of announcements and also their variety.

(e.g. the example of drones in Saudi)

The author figures this out by himself, which is great, but the "faster is better" (quote from the article) thing was a watering down of Boyd not the actual message. All things being equal, yeah, speed is better, but again that's not what Boyd was really saying.

This is a really important post, if you're the kind of person who thinks about this sort of thing. Iteration speed isn't the endpoint, just a measurement. Learning speed is more important, but only if you're also using that learning to identify and act on opportunities. All of the processes have their own notion of tempo or speed. You can benefit from thinking about them all.

Yeah, xingyiquan is like that.

But the half-beat is one of the more simpler, narrower application of OODA. Really, the whole thing with using OODA in an adversarial setting is that you can manipulate the adversary's OODA loop, of which the half-beat is one of those methods to do so.

I got exposed to Cynefin a couple months ago, and it got me thinking about how Cynefin and OODA are related. Probably the biggest thing is that Cyenfin is not necessarily about adversarial settings, and OODA can probably take different forms depending upon which Cynefin domain we are talking about. The ones that uses it in its simplist method (speed up the OODA loop) are probably working from the Simple domain, and perhaps the Complicated domain.

Boyd had shown a shortened form of OODA -- and this is where it starts intersecting with the likes of Musashi, or some of the old, obscure sword fighting manuals from Europe. What's not said on there is a form of intuitive information gathering that, well, frankly, people mind find dismiss as woo.

People sometimes talk about the OODA loop like it is a framework for action that you should adopt.

Another way to think about it is that the OODA loop is just the de facto framework of action.

---

You could also view this from the flip side: Is there a time when you might be vulnerable and you do not wish to be?

A trivial example: When I walk through a blind doorway or walk past a blind corner, I take a half step instead of a full step. I don't want to run into someone, and I don't want to anyone to run into me, so I step my back foot up to my front foot, but not past it. This gives me half a beat to evaluate the situation.

I do something similar at stoplights: I try to slow down as I go through an intersection, to allow more time for reaction.

> Another way to think about it is that the OODA loop is just the de facto framework of action.

Yes. The gains come from being able to go through the parts of the loop faster. Two classic Boyd examples from air combat in the Korean war were that US jets had fewer blind spots (Observe) and quicker controls (Act) [1].

[1] https://newbo.co/2017/10/25/scrum-fighter-planes/

The OODA loop reminds me of how a search engine works.

Observe - gather in trillions of web pages from every possible source.

Orient - filter those web pages until you get a small number that are related to a query.

Decide - analyze your top possibilities more deeply to get a ranked list of possibilities.

Act - render a web page with all the user interface conveniences to show to the users.

See also: Tempo timing, tactics and strategy in narrative-driven decision-making by Venkatesh Rao.

In the OODA model, acting at the wrong time just means you made the wrong decision. Waiting, changing tempo &c. are all actions, if done with intention.

Reply on news.ycombinator.com