Writers of Pro Football Prospectus 2008

Most Recent FO Features


» 2018 Free Agency Cost-Benefit Analysis

Is Kirk Cousins the best free-agent quarterback in recent memory? Should Trumaine Johnson or Malcolm Butler have gotten the larger contract? And what makes a free-agent contract good or bad, anyway?

07 Jan 2013

Week 18 DVOA Ratings

Once again, it is time for postseason DVOA ratings. As always, the following rules apply:

  • All 32 teams are ranked, whether they made the playoffs or not.
  • Teams are ranked in order of weighted DVOA, not total season DVOA. Since weighted DVOA is meant to lower the strength of older games, these ratings do not include Weeks 1-4, and Weeks 5-10 are somewhat discounted.
  • Teams which did not play in the wild card round are treated as if they had a bye week. (That includes both the 20 non-playoff teams and the four teams with byes.)

The playoff odds report is updated through the wild card games. You will find DVOA matchup pages for the four second round games on the FO Premium page. Remember that the equation used to determine win probabilities for the playoff odds report is not as complex as the one used for FO Premium picks, so picks may differ.

* * * * *

To save people some time, we remind everyone to put their angry troll hatred into the official zlionsfan angry troll hatred Mad Libs form:

<team> is clearly ranked <too high/too low> because <reason unrelated to DVOA>. <subjective ranking system> is way better than this. <unrelated team-supporting or -denigrating comment, preferably with poor spelling and/or chat-acceptable spelling>

If you are new to our website, you can read the explanation of how DVOA is figured here. As always, positive numbers represent more points so DEFENSE is better when it is NEGATIVE.

1 SEA 43.9% 1 12-5 29.5% 1 -7.8% 10 6.7% 4
2 DEN 39.4% 2 13-3 23.0% 3 -15.9% 2 0.6% 16
3 NE 38.3% 3 12-4 28.6% 2 -0.9% 16 8.7% 2
4 SF 28.6% 5 11-4-1 15.3% 5 -13.5% 4 -0.2% 18
5 GB 22.8% 4 12-5 16.6% 4 -7.2% 11 -1.1% 20
6 WAS 18.0% 6 10-7 13.9% 6 -3.6% 12 0.5% 17
7 CHI 15.2% 8 10-6 -10.3% 22 -22.0% 1 3.5% 7
8 CIN 13.4% 7 10-7 -5.2% 19 -13.9% 3 4.7% 5
9 CAR 12.3% 9 7-9 12.9% 7 -3.6% 13 -4.1% 30
10 BAL 11.3% 11 11-6 0.6% 14 -0.3% 18 10.4% 1
11 NYG 8.1% 10 9-7 9.0% 10 3.4% 22 2.5% 11
12 ATL 6.1% 13 13-3 4.4% 12 -2.7% 14 -1.0% 19
13 STL 3.7% 12 7-8-1 -1.8% 16 -9.1% 8 -3.6% 26
14 DET 2.8% 14 4-12 11.6% 9 6.9% 24 -1.9% 23
15 HOU 2.4% 19 13-4 -4.6% 17 -12.3% 5 -5.3% 31
16 PIT 2.4% 15 8-8 -8.4% 21 -10.2% 7 0.6% 15
17 DAL 0.1% 16 8-8 8.3% 11 11.4% 30 3.1% 9
18 MIN -0.3% 17 10-7 1.7% 13 5.3% 23 3.3% 8
19 NO -2.4% 18 7-9 12.2% 8 10.7% 28 -3.9% 29
20 SD -6.5% 22 7-9 -11.1% 25 0.0% 19 4.6% 6
21 CLE -6.8% 20 5-11 -14.2% 27 0.3% 20 7.8% 3
22 BUF -7.4% 21 6-10 -6.8% 20 2.4% 21 1.8% 13
23 MIA -10.5% 24 7-9 -10.6% 23 -1.2% 15 -1.1% 21
24 TB -11.2% 23 7-9 0.4% 15 8.8% 26 -2.9% 24
25 IND -13.7% 25 11-6 -4.8% 18 11.3% 29 2.4% 12
26 NYJ -21.2% 26 6-10 -22.9% 29 -8.3% 9 -6.6% 32
27 ARI -22.0% 27 5-11 -35.4% 32 -10.8% 6 2.6% 10
28 TEN -26.3% 28 6-10 -23.0% 30 -0.4% 17 -3.7% 27
29 OAK -27.5% 29 4-12 -13.5% 26 10.2% 27 -3.8% 28
30 PHI -29.1% 30 4-12 -10.7% 24 18.9% 32 0.6% 14
31 JAC -29.5% 31 2-14 -17.9% 28 8.0% 25 -3.5% 25
32 KC -38.9% 32 2-14 -24.7% 31 12.6% 31 -1.6% 22

The weighted DVOA ratings are slightly more spread out than the regular DVOA ratings, but we really go into the final eight without any surprising teams. In standard (not weighted) DVOA, the final four AFC teams are the four teams with the highest DVOA ratings, and the final four NFC teams represent the three teams with the highest DVOA ratings plus Atlanta, which was 10th overall but seventh in the NFC.

Below are the one-game DVOA ratings for the first round of the playoffs. The surprise here is probably that Green Bay and Minnesota aren't that far apart even though the Packers dominated that game from start to finish. It may surprise you to learn that the Vikings actually had more yards per play (5.2) than the Packers (4.9). The issue here is the fourth quarter, when Green Bay basically laid down and waited for things to be over. Going into the fourth quarter, the Packers had 35% offensive DVOA and -46% defensive DVOA. Then the Packers kept giving the ball back to the Vikings and inviting them to get back into the game, and the Vikings gave it a good try. The Packers had five drives in the fourth quarter and gained a net 17 yards with just one first down. The Vikings had the 35-yard drive that ended with a failed fourth down, then the long touchdown to Michael Jenkins, and a couple of long plays to end the game.

"But these plays were meaningless!" you say? Well, if the Packers were playing a better offense, perhaps they wouldn't have been so meaningless. It's not ridiculous to suggest that a team could come back from a three-touchdown deficit at the start of the fourth quarter. There is an element in DVOA that discounts plays in blowouts, but our definition of blowout for that variable is "over 21 points," and at no point did Green Bay have a lead of more than 21 points. Should we put something in to discount the final drive when a team is down by two touchdowns with 30 seconds left? Probably, but honestly, the amount that would improve the system's accuracy is miniscule. DVOA is designed to measure things in the long run more than in a single game, and those good Vikings drives in the fourth quarter would suggest that maybe this team was not as bad as it looked for the first three quarters. In Week 5, that would matter. When you lose in the playoffs, it doesn't matter, but that's what DVOA was trying to measure, and that's why the ratings for Green Bay and Minnesota are closer than expected.

DVOA (with opponent adjustments)
HOU 37% 10% -26% 1%
CIN -18% -21% 3% 6%
GB 33% 18% -16% -2%
MIN 11% -3% -8% 7%
BAL 44% 24% -11% 9%
IND -47% -13% 30% -4%
SEA 24% -1% -18% 7%
WAS 19% 0% -19% -1%
VOA (no opponent adjustments)
HOU 43% 7% -34% 1%
CIN -40% -39% 7% 6%
GB 29% 16% -15% -2%
MIN -6% -7% 5% 7%
BAL 67% 44% -15% 9%
IND -56% -17% 36% -4%
SEA 14% 1% -6% 7%
WAS -12% -12% -1% -1%

Posted by: Aaron Schatz on 07 Jan 2013

46 comments, Last at 28 Mar 2013, 5:01am by TedMoz


by dmstorm22 :: Mon, 01/07/2013 - 12:51pm

Was the table in the postseason DVOA columns always done with Weighted DVOA?

by buiscuits (not verified) :: Mon, 01/07/2013 - 12:59pm

Doesn't this conflict with Vince's espn article last week about how weighted DVOA is not a better predictor of playoff wins than regular DVOA?

by crw78 :: Mon, 01/07/2013 - 5:25pm

I was thinking the same thing. I think I also remember an comment in a forum or mailbag sometime within the last few years by by Aaron stating that Weighted DVOA was actually less predictive than Total DVOA for playoff games, but maybe I am misremembering.

by JIPanick :: Mon, 01/07/2013 - 6:11pm

I remember that too, but not where.

by Aaron Schatz :: Mon, 01/07/2013 - 7:08pm

Hey. I've always done the playoff DVOA listings with weighted DVOA. The reason why is that you don't need updated total DVOA, you already know what it is! (We've never done total DVOA that included both regular season and playoffs; it would take pretty strong playoff performances to move the numbers very much.)

by crw78 :: Tue, 01/08/2013 - 3:59pm

Yup, we can always go back to look at what the DVOA was after week 17 for the totals. Thanks Aaron.

by DisplacedPackerFan :: Mon, 01/07/2013 - 3:20pm

No arguments on the GB - MN ratings. I'm actually glad you gave the ratings for the first 3 quarters. Heck even Rodgers directly after the game mentioned the offense shutting down. Some of it is "classic McCarthy" 15+ point lead; run into the line, run into the line, throw deep on 3rd and whatever it may be, punt offense. There were execution errors though. How much of that was from over confidence, who knows. But it happens with the Packers regularly so I like that DVOA still considers it.

by Pat (filler) (not verified) :: Mon, 01/07/2013 - 3:29pm

Minnesota's VOA also gets a boost from the fact that Green Bay recovered that muffed punt, which cost the Vikings great field position and about 2 minutes.

The issue here is the fourth quarter, when Green Bay basically laid down and waited for things to be over.

I think the thing you should be stressing here is that Green Bay's offense basically shut down. It's one thing to say "okay, it's garbage time, we're going to give up yards" - it's an entire other thing to say "we're going to gain basically no yardage and burn no time off, either."

Green Bay's drives in the 4th quarter took off very little time, which is not a good thing for an offense to do.

It's also worth noting that the two teams were still 35% away in VOA. You'd expect something like a score of 25.2-17.7. Taking into account the muffed punt that's something more like 25-15 or so, and that's not too far off from 24-10 given scoring granularity.

by Nick Wells (not verified) :: Mon, 01/07/2013 - 11:49pm

Obviously Green Bay didn't finish the game the way it would have liked, but since Joe Webb obviously wasn't going to lead a three touchdown comeback, the outcome was effectively decided early in the third quarter, if not by halftime. I'm not saying that the Packers' inability to get a first down in the fourth quarter isn't a cause for concern, but DVOA would have us believe that the game was relatively competitive, when it wasn't at all.

by Andy Burdick (not verified) :: Mon, 01/07/2013 - 3:58pm

As a Bears fan, I'm curious -- how does the 2012 team compare to the highest DVOA ratings for teams not making the playoffs over the years?

by Thok :: Mon, 01/07/2013 - 7:43pm

Not particularly close: number one is the 2004 Bills with a DVOA of 31.3%. The 1991 Niners have a DVOA of 26.0%. There are some other teams I'm sure I'm missing.

by Aaron Schatz :: Wed, 01/09/2013 - 11:41am

The 2002 Chiefs were fourth at 24.4% but missed the playoffs at 8-8. It hasn't been this way for a few years, but the Chiefs are sort of the opposite of the Colts and Vikings, our numbers for a long time tended to rate them higher than their W-L record, particularly during the Vermeil years.

by Kal :: Mon, 01/07/2013 - 4:31pm

I'm really confused how Seattle's defense went down and their offense went up after the Washington game if the offense was at -1% and the defense at -18%.

by QCIC (not verified) :: Mon, 01/07/2013 - 4:53pm

Other teams had games too?

by Perfundle :: Mon, 01/07/2013 - 4:54pm

It doesn't look like their offense went up; the wDVOA went from 31.2% to 29.5%. I was puzzled by the defense as well.

by Arkaein :: Mon, 01/07/2013 - 5:45pm

Because these charts are Weighted DVOA, changes reflect older games that drop in weight in addition to the most recent game. If a defense goes down despite a good performance that might mean that there were some even better performances from a month or two ago that dropped significantly in their weighted contributions.

by Perfundle :: Mon, 01/07/2013 - 6:29pm

It depends what kind of weighting they do. The kind of weighting that would result in early games being dropped entirely doesn't sound like a good model, for precisely this reason. A better model would be exponential decay, as in
new wDVOA = 0.9*(old wDVOA) + 0.1*(last game DVOA)
This way, if the last game DVOA is better than the old wDVOA, the new wDVOA will definitely go up (though opponent adjustments and weighting might change this).

by Pat (filler) (not verified) :: Mon, 01/07/2013 - 11:10pm

The weightings are around somewhere, but they were calculated based on regression-testing to predict future DVOA.

I don't have any idea why a 'better' model would be exponential. The reason you reduce weighting on old games is because you assume that the team may be significantly different than 10 weeks ago, due to injury, player replacement, improvement, etc. Exponential decay would imply that there's some process that happens every week that alters a team. I don't see any reason why you would expect that a process like that would be smooth.

by leviramsey (not verified) :: Tue, 01/08/2013 - 12:51pm

Injuries are presumably randomly distributed (at least with respect the domain in weighted DVOA; if there's reliable information going back sufficiently far as to what games players get injured in, then maybe there could be schedule adjustments...). Lineup adjustments are also likely to aggregate to a random distribution. Improvement is likely to be a rather continuous process. All three of those factors should tend towards smoothness, barring analysis showing that they're not.

It's also an interesting question as to what's meant by future DVOA. Is it next season's DVOA? Is it post-Super Bowl (including postseason) DVOA? Is it DVOA for next week? Is it end-of-regular-season DVOA? All of those are useful definitions: next season would be great for the Almanac, but the post-SB and next week ratings are probably the most useful for predicting playoff performances. I could definitely see how the weightings would change between those definitions.

by Pat (filler) (not verified) :: Tue, 01/08/2013 - 11:57pm

All three of those factors should tend towards smoothness

No, actually, they wouldn't. Injuries toward significant portions of the lineup will be rare, but when they happen they'll persist for a while - several weeks. Ditto with lineup changes as well. That won't lead to a smooth distribution, because since they're rare, there's no chance for the random phase to smooth things out.

It's also an interesting question as to what's meant by future DVOA

End of season DVOA. Next season's DVOA is better predicted by full-season unweighted data, which is unsurprising since any mitigating factors (injuries) disappear by then, and losing data is always going to hurt predictivity.

but the post-SB and next week ratings are probably the most useful for predicting playoff performances.

DVOA isn't intended to predict the outcome of games. It's a balance between a predictive (correlation between seasons) and a retrodictive statistic (correlation to wins). Attempting to predict the outcome of games isn't actually that terribly interesting, because a huge portion of the outcome of games is nonpredictive.

by theslothook :: Wed, 01/09/2013 - 2:10am

Why do we assume injuries are normally distributed at all? I suspect injuries themselves follow a certain arch type model, highly heteroskedastic.

by allencp :: Mon, 01/07/2013 - 5:58pm

Seattle went from 18.5% weighted DVOA on offense last week to 29.5% this week after a -1% offensive performance?

edit: nevermind, that 18.5% last week was non-weighted.

by QB Eagles :: Mon, 01/07/2013 - 7:43pm

Someone needs to add the Legal Weed Bowl (i.e. Denver vs Seattle) to the special matchups section of the playoff odds report. Seems like the most likely outcome anyways.

by GroshKar (not verified) :: Tue, 01/08/2013 - 1:12am

Sea-Den: a.k.a. the "Why does this feel like an AFC West game" bowl

by Bobman :: Tue, 01/08/2013 - 1:32am

Just think: one of the first Seahawks games I went to after moving to Seattle in '96 was a Broncos game (they were more or less at their peak at the time) and if there weren't more Broncos fans than Seahawk fans actually at the game, it seemed like it and they sure were more vocal, rude, arrogant, and offensive. Lots and lots of orange in the stands and streets. Now the local papers are full of tales of outraged parents whose adolescent kids, wearing niners garb, were verbally abused by drunk Hawks fans at the week 17 game. (probably the same a-holes from '96 in new jerseys) What a fascinating world we live in.

by Kelib (not verified) :: Fri, 01/11/2013 - 8:05pm

That would be highest rated game of the year!

by Seth Lichtenstein (not verified) :: Mon, 01/07/2013 - 8:33pm

How could New England-Green Bay be a more likely Super Bowl than Denver-Green Bay if Denver is more likely to win the conference than New England?

by Cuenca Guy :: Mon, 01/07/2013 - 9:23pm

If Denver wins, Green Bay won't be as likely to win because they'll be depressed thinking about the Super Bowl they lost to the Broncos with John Elway.

by Perfundle :: Mon, 01/07/2013 - 9:46pm

Well, it's a simulation. It's not the exact percentage calculated with each team's DVOA numbers. If they did the simulation again, it's possible one or both of those scenarios flip.

by Pat (filler) (not verified) :: Mon, 01/07/2013 - 10:59pm

Because in the simulations where Green Bay won the NFC, New England won the conference more times than Denver did. These two things are (probably, in the simulation) unrelated, so it's just statistical noise.

I say 'probably' because you *could* make the simulation so the probabilities are slightly joint: if you did something like "assume Green Bay is slightly stronger than we think they are" (i.e. you pick an 'error' for each simulated run) and then back-calculate the 'most likely' new DVOAs from opponent-adjustments using your new knowledge of Green Bay. That could make the probabilities slightly joint... but it's almost certainly just statistical noise, because I'm pretty sure they don't do that.

by intel_chris (not verified) :: Wed, 01/09/2013 - 1:52pm

I was just reading the playoff odds report and noted that same anomaly. It probably is a simulation artifact. I have no insight into Mike Harris's simulations so I won't besmirch them. They are certainly good enough for entertainment value. It does give one an approximation of the odds of the different events happening.

It would, of course, be nice to have a way of calculating for two teams with DVOA's or X and Y what the odds are of each team winning (with error bars). Still, what we know from DVOA is that one single number doesn't represent well any given team. So, one would do better using the offensive, defensive, and ST DVOA's. Or, better yet, ones that take even more nuances into account. Perhaps Mike's numbers actually simulates down to the levels of individual plays and uses the DVOA numbers for down and distance. That would explain the need to do such a large number of simulations.

However, whatever the simulation does, there is a certain noise and uncertainty in the prediction (the error bars) and we don't know that. What we can gather from the anomalous predictions is that that uncertainty is large enough to cause entries in the table the seem contradictory. (I suspect the error bars are even larger than that.) From that one can infer that teams which are close in odds in the table are probably close in odds to achieving the listed outcome. Thus, Denver might be slightly more likely to advance to and win the SB than NE, but the two teams are probably close enough that the variability of any given football game makes such predictions a crap-shoot. (Or as they say, "that's why we play the game.")

Still, I want to challenge one thing which has come up in the thread as it has in others. DVOA can change as we get more information. Teams sometimes are and sometimes aren't who we think they are. More information gives us a better idea of who they are. However, a simulation based upon DVOA cannot realistically cause us to adjust that value up or down. Sure the random numbers can cause different teams to win. But if the simulation is modeling the statistics properly, the DVOA of the teams simulated--i.e. what their tendencies are should not be affected. Those tendencies don't change just because we throw dice and they come up with different numbers. The simulation hasn't added any information. It has just used that information to make a prediction. Thus, adjusting DVOA based upon that prediction would be wrong, because we would be adjusting DVOA based upon random factors, not increased knowledge.

Thus, while the odds of a particular team making the SB might be +/- 10% (or some other figure) and thus the odds of a particular matchup be +/- 20%, which would swamp the odds predicted. That is just the uncertainty in the odds, and trying to backfill the odds from the simulation results to get more precise odds doesn't work. We have to live with the fact that we don't know how the games will play out, and enjoy our predictions and hope the ones we like come true and the ones we don't are off.

by Paul M (not verified) :: Tue, 01/08/2013 - 1:27am

If McCarthy had wanted to, the Packers could have scored at least two more times and won by 30+ pts. You see, it's a funny thing-- he don't care about DVOA. I wasn't happy as a Packer fan to see them shut down, and the playcalling become so pedestrian-- but nobody got hurt and it's on to SF.

by Pat (filler) (not verified) :: Tue, 01/08/2013 - 10:38am

If he really wanted to make sure no one got hurt, he wouldn't've been calling multiple passing plays on the drives in the fourth quarter. Including one where Rodgers was sacked.

by Paul M (not verified) :: Tue, 01/08/2013 - 1:32am

"If the Packers were playing a better offense, perhaps they (the bad 2nd half offensive plays) wouldn't have been so meaningless."

Disingenuous at best; outright disregard for the facts on the ground at the worst;

Aaron, do you think for a second vs. Brady, Manning, Kapernick, Wilson or even Ryan in the Georgia Dome (and maybe expecially the latter) that Mccarthy shuts the offense down after gaining a 3TD lead on the first drive of the second half?? Of course not-- he knew Webb couldn't beat them-- so he chose not to risk a turnover or an injury-- I can't really fault him much-- but more importantly I wouldn't waste any space trying to defend an obvious flaw in DVOA which this situation exposes.

by Ender (not verified) :: Tue, 01/08/2013 - 10:42am

"Aaron, do you think for a second vs. Brady, Manning, Kapernick, Wilson or even Ryan in the Georgia Dome (and maybe expecially the latter) that Mccarthy shuts the offense down after gaining a 3TD lead on the first drive of the second half??"

One would hope he wouldn't but he did exactly this in the Colts game this year and that is the biggest reason they lost that game. He does have a history of playing overly conservative with a big lead in the 2nd half.

by B :: Tue, 01/08/2013 - 9:25pm

Isn't the real issue that the Packers can't run the ball, especially in obvious situations, and this makes it hard for them to protect a lead?

by Ender (not verified) :: Wed, 01/09/2013 - 11:45am

Yeah that is definitely a concern. Some teams can play conservatively and still move the ball. The Packers can't really kill a game on the ground the way many teams can.

The inability to run on 3rd and 4th and short is a big weakness too of course.

by Anonymous1 (not verified) :: Tue, 01/08/2013 - 8:58am

DVOA is more impressed by Houston's win than any fan or analyst seems to be.

by Ender (not verified) :: Tue, 01/08/2013 - 10:37am

The Packer DVOA doesn't surprise me at all. By the middle of the 3rd quarter they shut down the offense and went run, run, try to convert on 3rd down. If it had been a close game they probably score 40+ but all they did for the last 1/3 of the game was milk the clock since the Vikings couldn't put points on the board.

by Ender (not verified) :: Tue, 01/08/2013 - 10:38am

Btw, this is also why the Seahawks #1 DVOA on the year doesn't mean as much to me as it should. They ran up points in a lot of situations where all the other teams shut down their offense.

by Insancipitory :: Tue, 01/08/2013 - 10:04pm

You should watch the 4th Quarters of the Dallas and Vikings in Seattle to see what DVOA loves about them (aside from opponant adjustments). They didn't run run try punt, repeat. They went run fling run run, repeat, repeat, repeat, repeat, kneel.

People seemed to be wrapped up in the idea that DVOA should be WPA or something like it. The Packers weren't trying to kneel, kneel, kneel punt, but DVOA would probably respect that tremendously if the Packers got booming punts with insignificant returns.

I would also add, that in the past few weeks the Seahawks offensive playcalling at the end of games seems to be more erratic. It's almost they're using those moments as opportunities to use the other team to scout themselves. They have a tremendous amount of confidence that neither Wilson or Flynn or going to make a bad mistake to bring their opponants back into it, and are completely willing to gamble/experiment accept the result and put it all on the defense. Earlier in the season, they didn't do that. They were agressive (still going for it on 4th down) but conservative. And to see them execute it was pretty brutal; violent even.

It could be that's what the Packers are doing too, but when it doesn't work on the field it's unsuccessful. If it proves successful in film study, it shows up in the next week's DVOA.

by Jetspete :: Tue, 01/08/2013 - 10:45am

the best part of that table is how it shows just how dominant houston was against Cincy, except the Bengals came within an Andy Dalton overthrow of winning the game

by maxnote :: Wed, 01/09/2013 - 2:46am

How come Atlanta's weighted DVOA is 6.1%, when at the end of the regular season it was only 2.4%? Does weighted DVOA use bye weeks as "weeks" to decrease the weight of games from 8-12 weeks ago?

by nath :: Wed, 01/09/2013 - 6:32pm

Is there a way to get week-by-week DVOA graphs or is that only available with a Premium subscription (or when FO feels like charting a team)?

by okalriii (not verified) :: Thu, 01/10/2013 - 7:22pm

gonna go out on a limb and say either Green Bay or New England is you're Super Bowl champion. How do I know? 8 of the last 9 teams that have won the Super Bowl finished in the Top 10 of the previous year end rankings for Points Per Game AND were in the Top 10 Average Scoring Margin for the Past 3 Games at least 3 weeks before the Super Bowl. This year's members are GB and NE.

by TedMoz (not verified) :: Thu, 03/28/2013 - 5:01am

Seattle is the best team i ever see. Players are in good mood and more than aggressive than in the past season. They will get the champion CUP i feel this!
my site: http://www.bestcustomwriting.com/deductive-essay