analytics

Falling Victim to the Whims of Chance

I did an analysis of my Vancouver photographs looking at streaks of outings with no first-rate photographs.  The streaks were longer than I expected, but that can be attributed to the whims of chance.  My analysis of streaks re-enforced for me the importance of continuing to take photographs even if there has been a prolonged period of disappointing results.

For more information on methodology see my post on Photography Analytics.

I analyzed 101 Vancouver outings.  There were 30 outings with one or more first-rate photographs for a probability of 0.3 (30/101).  That's consistent with what I expected.  Before doing the analysis, I estimated 1 in 3 outings resulted in at least one first-rate photograph.  

The number of first rate photographs on any given day depends on chance.  It’s like flipping a coin.  I know I’ll get heads or tails on each flip.  I also know that I might get several heads or several tails in a row.  With the photographs, I know there’s a 1 in 3 chance I’ll get 1 or more first-rate photographs on any outing.  There will be streaks of consecutive outings with first rate photographs.  There will also be streaks where there are no first-rate photographs for several outings in a row.

Our intuition often fails us when it comes to randomness:

“People tend to think that streaks in random sequential events are rare and remarkable. When they actually encounter streaks, they tend to consider the underlying process as non-random.”

from http://www.ncbi.nlm.nih.gov/pubmed/20728080

This psychological expectation can play havoc with my confidence.  Based on my estimate of 1 in 3, I accept not every outing will be successful and that unsuccessful outings will outnumber the successful ones by 2:1.  It’s natural to expect luck to be evenly distributed.  That is, there will be a successful outing followed by a couple of unsuccessful ones followed by another successful one and so on.

Chance - randomness - doesn’t work that way.  Rather than being the exception, streaks are the norm.  

Even knowing this, I still expect - or at least hope - my luck will be fairly evenly distributed.  I start to feel a little hard done by if there are more than two days in a row without any first-rate photographs.  If the streak extends to three days or more, it begins to feel like a mounting crisis.  On the other hand, sometimes the wins come more frequently.  Rather than attributing this to good luck, I start to believe I have somehow “levelled up” and will now continue to produce at a higher rate.  It’s an emotional see-saw.

Here are my streaks of consecutive outings where I had no first-rate photographs.  A streak-length of zero means I had one or more first-rate photographs in back-to-back outings.

101 outings, length of streaks with no first rate photographs

My streak lengths have a median value of 2 and 75% of the streaks are three or less outings in length.  But 25% of the streaks are 4 or more outings in length and I went on one streak lasting 8 outings in a row!   That can seem like an eternity.  If you're only going out once a week that two month stretch might be enough to convince you to pack it in.  (it feels a little less painful when I'm shooting more frequently than that.)

It's tempting to think I must have been doing "something wrong" during those streaks.  However, this distribution looks similar to randomly generated streaks.

I built a simulation of 101 streaks where the probability of 1 or more first-rate photographs in an outing is 0.3.  Here's the results from three different simulations:

The streaks vary considerably with every simulation.  

It turns out a streak of length 8 or more is quite likely.  I ran the simulation 10,000 time recording the longest streak of each simulation.  Using these simulation results, I calculated the probability of a simulation having at least one streak greater than a given length.   

10,000 simulations of 101 outings with probability of one or more first-rate photographs in an outing = 0.3

100% of the simulations had a streak at least 4 outings long and 69% of the simulations had a streak of 8 outings or longer.  Rather than being an exception, a streak of 8 is to be expected. 

The expected length of streaks is dependent on the percentage of outings with one or more first-rate photographs.  For example, here's what the simulation looks like if the probability increases to 0.5.

10,000 simulations of 101 outings with probability of one or more first-rate photographs in an outing = 0.5

Even if the odds are 50/50 of getting a first-rate photograph on an outing, there is a greater than 50% chance of having at least 5 outings in a row with no first-rate photographs.  Ouch.

I conclude that despite my best efforts, streaks will happen.  Rather than getting discourage I just need to accept it and carry on.  

The photograph that preceded the streak of 8 consecutive outings without any first-rate photographs.

The Infatuation Half-life

Based on my selection process over the past month, my infatuation half-life for new photographs seems to be about two weeks.

“Infatuation is the state of being carried away by an unreasoned passion or love.”
— http://en.wikipedia.org/wiki/Infatuation
“Half-life (t1⁄2) is the amount of time required for a quantity to fall to half its value as measured at the beginning of the time period.”
— http://en.wikipedia.org/wiki/Half-life

I haven't been sharing those photographs because I decided to slow the sharing pace.  But, I'm photographing regularly and making selections as I go along.

I use Lightroom to manage photographs.  After an outing, I do a first pass and mark the ones I like with a five-star rating.  I'm generous at this stage, giving photographs the benefit of the doubt.  After a day or two, I revisit the outing and revise my ratings.  I may add a five-starting rating to some and decrease the rating of others.  I'll add the ones with a five-star rating to a project-specific collection.

Within a collection, the first-rate photographs have a five-star rating.  The newly added photographs join that esteemed collection.  Some may be down-graded right away when I see them in the context of other first-rate photographs.  

Often, there are new ones I'm sure will make it as first-rate photographs.  I mean, positively certain.  There may even be several like that - which is very exciting.

I review the collection from time to time, as a minimum whenever I add new photographs to it.  As I do this, the ratings of recently added photographs begin to fall.  The ones I thought were definitely most excellent, fade to a more provisional status.  On the next viewing, it becomes clear most of the provisional ones definitely don't warrant a five-star rating.  Those infatuations are over, replaced by new ones.  That process of going from "almost certainly a first-rate photograph" to "definitely not a first-rate photograph" takes about two weeks.  The ratings seem to stabilize after that.    

Recently, I've discovered a mental test that may be shortening the infatuation half-life.  I ask myself: "When I decide to share these new photographs, would I select this as the first one to share?"  If the answer is no, I downgrade it.  

I think there will be a longer term "competition" to keep those five-star ratings, driven by factors other than infatuation.

Iona Beach, 2013

Photography Analytics

As a product manager I make extensive use of analytics to gain insight and make decisions.  I'm working on some posts discussing photography analytics.  In this post, I want to discuss my motivation and methodology.  

I photograph with the hope of creating "first-rate photographs".  They are first-rate for me and may not be for other people.  They are the ones I'd include a portfolio.  They are rare and my choices are somewhat fluid with photographs falling in and out of favour.

I believe luck plays a role.  Therefore, persistence is required for success.  

The need for persistence is not a profound insight.  I hope analytics will provide further motivation for persistence, by quantitatively re-enforcing it's value!  

For example, I'm currently looking at streaks.  My hypothesis is that dry spells happen and that chance alone can explain a lot of it.  The best strategy is to just keep photographing.

I work in "outings".  That is, I pick up my camera and go out with the intention of making photographs for inclusion in a portfolio.  During an outing I'll make some number of "exposures".  If I'm lucky, one of those exposures will be a first-rate photograph.  If I'm really lucky, more than one exposure will be a first-rate photograph.  But (in my case) there are no first-rate photographs from the majority of outings.

Most of my work is digital and I store work by date so keeping track of outings is straight-forward.  Some while after an outing, I can record the number of first-rate photographs.  The Prairie Modern photographs were made with a 4X5 film camera, so I won't do analytics for them.   

People using analytics are fond of using acronyms such as FRP for first-rate photograph.  I'll try to avoid them, but may resort to using a few if I find it getting repetitive.

An example of a statistic is the average number of first-rate photographs per outing (0.4 for my Vancouver photographs).  

Another is the percentage of outings that have one or more first-rate photographs (0.3 for the Vancouver photographs).  That's also the probability of one or more first-rate photographs from an outing.

(The probability of having one or more photographs in an outing is lower than the average number of photographs per outing because some outings have more than one first-rate photograph.)

Hopefully that wasn't too bruising.  I should publish the first photography analytics post next week.

One of the earliest first-rate photographs:  

Sunlit Trees, Medicine Lake, Jasper National Park, 2002