[arishi1]

Smartcat thanks for including Optra 2.0 TDCI in this listing. I was expecting the high scores from Optra 2.0 TDCI, thats the biggest diesel in this category. Now lets see how it scores in the Fuel Economy Section.

[LithiumSunset]

I am in the hunt for a sedan and your thread is so timely..

[Dr.AD]

Great analysis and great work, Smartcat! We really appreciate all the hard work you have put into this analysis. Also, your idea of creating a quantitative scoring system is great. Thanks for the great work. I have also followed your similar analysis on hatchbacks and really enjoyed that thread.

If you don’t mind, I would like to suggest a few modifications to your calculations to make the comparison even more meaningful. I was going to suggest these modifications for your analysis if small cars, but somehow forgot to put my thoughts down. My suggestions are:

1. The point system of giving 25 points to the top car and 1 points to the bottom car is a bit unfair. The bottom car may be very close to the top car in a given category (e.g. performance) and may not deserve such harsh penalty. Similarly, in-between cars need not all get 1 point penalty for one rank regardless of their actual performance. This point system does not take into account the actual value of the parameter being considered (e.g. 0-100 time, 30-80 times, FE etc.), and this is also not “normalized”. Therefore, I would like to suggest a “normalized” point system as follows:

1.1. Let lower_limt be the worst case value of the parameter for all cars under consideration. For example, for 0-100 performance, lower_limit = 17.86s (corresponding to Mitsubishi Lancer 2.0 Diesel )

1.2. Let upper_limit be the best case value of the parameter. For example, upper_limit = 9.06s (Skoda Octavia 1.8 Petrol) for 0-100 performance.

1.3. Let us define a

normalized goodness index = (actual value-lower_limit)/(upper_limit-lower_limit)

For example, for Fiat Linea 1.4 Petrol (15.14s), the normalized goodness index for 0-100 performance will be = (15.14-17.86)/(9.06-17.86) = 0.31. It is easy to see that this value will be 1 for the car with best value, and 0 for car with worst value. Thus, this is both normalized (on a 0 to 1 scale) and considers actual value and not just the rank position.

1.4. Use this normalized goodness index instead of your point system.

The real use of normalization will be clearer after the 2nd point below.

2. How to we compound the performance of a car in different categories? You have used simple addition of scores. Addition is statistically not the best method. Addition has the bad property of cancelling high and low values and not differentiating between “all average” case versus “combination of great and terrible cases”. For example, let us consider two categories A and B. Let us say in your system, Car 1 scores 10 and 10 in both categories, whereas Car 2 scores 18 and 2 in these categories. You would give them both scores of 20 (by addition), and this completely misses the fundamental difference between these cases. Therefore, I suggest the product of the scores. Statistically, if all quantities A, B, C etc. are equally desired, then the net quality is (AxBxC), and not (A+B+C). With products, the score for Car 1 in above example will be 100, and that for Car 2 will be only 36. Thus, really bad performance in any category is penalized and more balanced performances in all categories are rewarded. Of course, great performance in all categories (if there is such an example) will be highly rewarded.

2.1. Now, you can see that the products (AxBxC) can fluctuate between very high numbers (e.g. 26x25x26 etc.) to very low numbers (e.g.26x0x20 etc.). This fluctuation is because of non-normalized numbers. This is where normalization that I mentioned above comes into picture. Using the normalized numbers above and then taking the products (since now all are on uniform 0 to 1 scale) makes statistically better sense.

2.2. It can be easily seen that if a car comes last in any one category, it gets 0 in that category and hence the net product is always 0, regardless of scores in other categories. Thus, this system eliminates worst case performers in any category. One needs to have at least some decent performance in all categories to get decent overall score.

2.3. I mentioned the point “equally desired” above. If the parameters are not equally desired, then one can easily add “weights” to them. You already mentioned that you are assuming equal weights, and I agree with this assumption, and hence I think we can stick to equal weights.

2.4. The most accurate way of compounding these scores (over different variables) is to calculate the area of a polygon, where each axis of the polygon corresponds to normalized score for that car in each category, with origin at (0,0,0…) for each category. In other words, if you plot a “radar plot” in Excel with the normalized score, you will get this polygon. The reason this is most accurate is that even if the car scores 0 in any one category, the net area is substantially reduced, but it is not 0. However, this is a little bit complex exercise; hence I suggest products as a simpler method.

Please let me know what you think about these suggestions. Please feel free to let me know if I have missed something of if this does not make sense.

Once again, thanks for the great work!

Regards,
-AD

[pnredkar]

Nice thread. Keep up the good work.

I remember that you had asked about how to update the rankings automatically. You can use the "RANK" function in Excel. I have attached an excel sheet that illustrates this formula.

Let me know if you need further excel help.

- Prasad.

[ACM]

Dr. AD and Smartcat.

Nice suggestion on Normalizing the values. But But this makes it complex inspite of being fairer. Though unfair, I would still recommend that Smartcat atleast for now goes ahead with the analysis the way he is going it. It keeps the process transparent. (Like he has given the Magazines the credit for all the data (irrespective of it's errors) and is resisting the temptation to correct some.

But the normalization method could also be carried out in parallel. There is a major value to it as well.
The reality is that in todays times yes 11 sec is a real good time for a budget sedan (0-100) but then anything above 16-18 sec in a sedan is really a criminally underpowered car. (though in many SUV's there are good times.) So a "0" for the slowest car is not really out of sync with reality.

Yet the normalization suggestion also has 1 major benifit that if there were 1 car say at a high speed etc parameter and all the rest were close to each other near the lowest value or for that matter heighest value then their scores get normalized.

I may be best if you could do an alternative calculation based on your system of normallization and post back to this thread the calculations with the normalization method.

The Thread is progressing brillantly, the Smartcat appreciate the service to the forum, with the kind of time you are putting in. The suggestions by followers of the thread are also very impressive.

Once done the entire process could be documented without post like this one and then reposted continuously. There is a market for that as well.

[lifeonwheels]

Excellent thread Smartcat

The style of ranking seems to be reasonable and the results acn't said to be far from reality. Chevrolet Optra was an intersting inclusion and it did knock a few cars off their podium places. It certainly deserved a place since it retails for less than 12 lakhs due to heavy discounts.

Regarding the figures I do accept that testing vehicles would have been better but then again it doesn't turn out to be a practical option so going by the numbers of one magazine for all cars is the best possible method. Keep such great and informative threads coming through.

@Gilead - There's a lot more to a car's handling than the ACI track test in which the Honda CRV surprisingly fared better than Cedia and Civic as well!

Interesting article below from Overdrive.




The ACI article says that Fiesta is the best handler and this one says Cedia is better than Fiesta so which one do we believe. Either both cars are useless in trems of handling or both the articles are useless. The point I'm trying to make here is that Fiesta and Cedia are both great handling cars. Handling is something which a user should judge for himself and just dragging along one article frequently doesn't make much sense. The reason for putting up this article is not to prove Cedia or any other car superior but to prove that articles solely do not prove anything coz if they did then which one do people believe?

Sorry for the bad picture quality. In case the text is not clear please use the link below and the above mentioned page is No.24.
http://issuu.com/infomedia18/docs/inrc_booklet_web

[benzinblut]

@smartcat: thanks for including OptraMagnum TCDi. Im happy now and even more interested.

I thought the mangnum had a better 0-100 timings.

for louzy 30-80 and 40-100, turbolag is the culprit. Take the runs in a gear lower and it will give you far better timings apart from the headrush! (especially 30-80 in 2nd.. whoa!!)

[Gilead]

lifeonwheels, can you confirm if the rally Cedia was running a stock set up? The article says the Cedia was running on Evo parts and high lift cams and ITBs. Is it any wonder that its 200 BHP engine could outperform a Fiesta? And I bet it's suspension was far stiffer than stiock. All cars were running stock in the ACI test if you recall. Let's stick to discussion of regular stock cars here and not tuned up versions where the final outcome depends on the tuner's ability. And in stock form, the Fiesta is indeed faster around a track than the Cedia.

[lifeonwheels]

Gilead, if the Cedia was running on high performance parts I'm sure there's no way the Fiesta was running in stock form. Also the Fiesta was already three races old when Cedia made its first outing so the Fiesta would have surely had more tuning time.

A highlighted para clearly speaks about only rally tyres being shod onto the Cedia and it still performed very well. No changes to engine or suspension then.

Yes your talk about the stock thing is right and I never stated in my post about one car being faster than the other. All I said was about articles not being the final verdict since every race driver has a uniques style of driving. For eg most drivers do not prefer cars which oversteer but Michael Schumacher prefers that. So a car which might be faster in the hands of MS on the same track might be slower than the other on the same track. One more highlighted para clearly states that initially when not modified the Cedia was slower than Fiesta in speed. This clearly shows that the Fiesta was modified and the Cedia wasn't and still the Cedia won because we all know that a stock Cedia has a higher top speed than a stock Fiesta.

My point is just that a road test by a magazine does not prove anything and it shouldn't be always considered as a benchmark. I know you are happy with your car and I'm happy with mine and one being faster than the other is a point of very little or no relevance.

[vasoo]

Excellent work smartcat. By the way, the in-gear acceleration listed in ACI website is actually for 20-80 kmph in third gear and not 30-80 as mentioned in your analysis. Please correct the same.

[Gilead]

lifeonwheels, it says that the Cedia was tuned to over 200 BHP. I am sure the Fiesta was tuned as well, but we have no idea to what extent etc. That's why I said let's compare stock timings of both cars driven by the same person. We have timings from Gaurav Gill and Narain Karthikeyan and the Fiesta was faster in their hands.

I am happy with my car and I am sure you are happy with yours as well. I did not even mention the word Cedia before you popped in with that article. I agree that timings have no relevance in real life, but they sure do in a thread comparing performance of cars. And besides, the Cedia is leading the ranking as of now, so you should be happy.

[lifeonwheels]

Gilead.. the 200 hp car is different from the one lagging in terms of speed.

Yes figures do have their relevance in such threads and highlighting their limitation was my sole aim.

I'm sure if Honda would have still sold the base model Civic it would have found a podium spot in this list.

[arnabchak]

Smart Cat...

Thats a wonderful analysis.

The Verna Diesel coming on tops among petrol heads is a surprise.

The Fiesta Diesel languishing towards the bottom of the table is also pretty surprising!!

[Gilead]

Originally Posted by lifeonwheels

I'm sure if Honda would have still sold the base model Civic it would have found a podium spot in this list.

I am not too sure about that. The one Civic I have driven till now had very poor low end torque. I don't know if that was an issue with that particular car or whether that's how all Civic engines are tuned. My Fiesta has much better low end pull and I won't be surpised if the Fiesta and Cedia have better in gear acceleration than the Civic. It will be a different story for a 0-100 drag or outright top end.

smartcat, what are the parameters you will be using in the next few rounds? Give us a sneak peek. EDIT: Just read your opening post again. It will be interesting to see how you assign rankings for features.

[lifeonwheels]

yes Gilead - Civic doesn't sport a good low end torque but in the 0-60 and 0-100 timigs it would have been ahead of Cedia marginally and maybe at par with Skoda. The in-gear timings are certainly something I will scout for although I'm sure Fiesta and Cedia would fare better here but then again curiosity gets the best of me mostly