Its a bit daft to say AI doesn't matter because you don't personally agree with the results. It does matter. The results are collated scientifically and in a manner that is far more representative of viewers than this forum.

For many people - viewers, not fans - Doctor Who had reached a pinnacle by the end of season 4. I know that's not a popular view here (particularly among those who prefer Moffat's version of Who), but the AI figures reflect this.

The show is still doing well, however, and I am personally surprised Cold War got only 84 but then I am a Who fan, and the AI reflects viewers more generally.

I think the AI figures are important although I am a little surprised at how there isn't a great deal of difference between a high figure of 91 say and a low figure of 84. I guess those in know, know better than me how to interpret these things but it would be easier for us joe blogs to interpret if they were wildly different.

I agree, but i can't help thinking that it should be doing a little bit better.... That's just me.

The fan in me wants AI's in the mid 90's and overnights in the mid 8m!

I would like an AI of 100. Nothing less is acceptable to me Come to think of it in the whole history of TV has any programme ever achieved an AI of 100? Surely that must be impossible!

Complete nonsense.

Your first 2 paragraphs merely confirm your personal opinion about which episodes deserve a high rating and which don't, they mean nothing to anyone else.

The AI takes 5,000 viewers opinions into account so is by far the largest and most unbiased official calculation of an audiences general views about an episode.

You may not agree with them, but to say they don't matter because episodes you personally don't like are rated higher than ones you do is rather egotistical IMO.

Its a shame to say, but I think a lot of Who fans would agree with Josh Pinder. I am always very much aware that the show has to appeal to the biggest audience possible if the BBC is going to continue investing in it/making it. Sadly, I think a lot of Who fans want a show that appeals to them first and foremost.

I think Moffat allows his own 'inner Whovian' to affect his take on the show and it does feel as though it has become more insular than it was during the RTD/Tennant/Ecclestone years.

Do not rely on one statistical measure only. Second there is a probability (which I am assuming is low but it is not zero) of the high and low AIs occurring due to chance so do not treat the estimates as the gospel truth.

Also Doctor Who is becoming more popular worldwide especially in the US so I wouldn't say Doctor Who is appealing at a narrower audience.

You're right about not relying on one statistical measure. The AI is scientific and representative so isn't the result of chance. And Who's been big in the US for decades and is no more popular now than 40 years ago ( when it was watched by 100m people worldwide).

Hi Yog101. do you have the figures for The Voice and BGT (exc and inc +1) please? If you could post them in the main Ratings Thread I'd be grateful!

It can result from chance, reading their methodology on Wikipedia hasn't changed this view. It is only an estimate from the sample you cannot be 100% sure an AI of 91 is the true population value. A good estimator can still produce inaccurate estimates (assuming the variance isn't 0) They certainly haven't asked my view on Doctor Who episodes. Also I disagree with you and maintain my view that Doctor Who in the US is becoming more popular.

But it's exactly the same as ratings, not every single viewer is counted, a select number are counted and the total extrapolated across the whole of the nation.

Of course it's not 100% accurate, such a thing is a logistical impossibility, but it's the best way we have to calculate ratings, just as the AI index is the best indicator we have of an audiences view of a particular episode.

A typical opinion survey of 1000 people (weighted so its representative) has a margin of error of +/-3% so for 5000 its going to be even smaller.

It's also worth noting that a margin of error (which is measurable) is not the same as 'chance' (which is not).

The point I am making, however, is that a viewer can't chuck AI out of the equation because they didn't personally like the story that delivered the highest AI rating in Who history.

(Thanks to Andy Parish at Gallifrey Base.)

You can measure chance, assuming the AI score for each person is independent of each other and there is a large sample the mean of the AI score will approximately follow a normal distribution (central limit theorem, please correct me if I am wrong) therefore you can measure probability by integration (using z tables in this case).

Also I am guessing for that margin of error they said there is an extremely large probability the true value is in this interval so that there is a very small probability (almost zero) of it being outside that interval.

All I am saying is that when comparing doctor who episodes you cannot say for certain that one episode is better than another based on AI scores especially if they are close. You certainly can't make claims about one era having a broader audience than another based on close AI scores alone.

I think you've misunderstood what an AI rating is. It doesn't measure what's good but what's liked. And like it or not, the AI rating is an industry standard and does have impact. If it didn't then it wouldn't be used.

Fine then replace better with more liked then. It doesn't make any difference to the point I am making.

Cold War update: 6.931m

Should break 7m tomorrow



I shall do just that, later today
(providing I have the time)

Are we in line for a sub 1.5m time-shift. When was the last time that happened.... If ever?

The point is, scientifically, the AI rating is a very precise recording of an audience reaction to an episode and the differences (though small in some cases) are robust.

It doesn't measure quality, just simply whether the audience enjoyed it and (if so) to what degree.

Personally, I'm surprised Cold War got the same rating as the previous episode as I thought it was much better. But the AI rating doesn't reflect my opinion (or the opinions of Who fans) but the audience more generally.

And its that more general audience that needs to be kept happy if Who is to continue getting made. That's why the AI is an important complimentary measurement to the viewing figures.

Night Terrors had a timeshift of around 1.46m, that was the last time it was under 1.5m

Excellent post

No it isn't. It doesn't change anything unless you are trying to claim the variance of the AI estimator is zero? I was about to post that it was complete waste of time posting it. You can get high or low AIs due to chance. I am not wasting any more time discussing basic statistics.

Night Terrors was just under 1.6m wasn't it?

5.5m > 7.07m

thank you!

Night Terrors had a timeshift of 2.88m, from 4.19m live to 7.07m final figure. (This includes VOSDAL.)